PLM

Package ‘plm’
December 2, 2016
Version 1.6-5
Date 2016-12-02
Title Linear Models for Panel Data
Depends R (>= 2.12.0), Formula (>= 0.2-0), stats
Imports MASS, bdsmatrix, zoo, nlme, sandwich, lattice, car, lmtest
Suggests AER
Description A set of estimators and tests for panel data.
License GPL (>= 2)
URL https://cran.r-project.org/package=plm,
https://r-forge.r-project.org/projects/plm/
Author Yves Croissant [aut, cre],
Giovanni Millo [aut],
Kevin Tappe [aut],
Ott Toomet [ctb],
Christian Kleiber [ctb],
Achim Zeileis [ctb],
Arne Henningsen [ctb],
Liviu Andronic [ctb],
Nina Schoenfelder [ctb]
Maintainer Yves Croissant <[email protected]>
Repository CRAN
Repository/R-Forge/Project plm
Repository/R-Forge/Revision 371
Repository/R-Forge/DateTimeStamp 2016-12-02 10:18:21
Date/Publication 2016-12-02 23:31:22
NeedsCompilation no
1
2 R topics documented:
R topics documented:
plm-package . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
aneweytest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Cigar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
cipstest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Crime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
detect_lin_dep . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
dynformula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
EmplUK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
ercomp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
fixef.plm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Gasoline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Grunfeld . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Hedonic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
is.pbalanced . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
is.pconsecutive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
LaborSupply . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
lag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
make.pbalanced . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
make.pconsecutive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Males . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
model.frame.pFormula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
mtest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
nobs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Parity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
pbgtest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
pbltest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
pbsytest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
pcce . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
pcdtest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
pdata.frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
pdim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
pdwtest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
pFormula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
pFtest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
pggls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
pgmm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
pht . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
phtest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
piest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
plm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
plm.data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
plmtest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
pmg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
pmodel.response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
pooltest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
plm-package 3
Produc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
pseries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
punbalancedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
purtest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
pvar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
pvcm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
pwaldtest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
pwartest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
pwfdtest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
pwtest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
r.squared . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
RiceFarms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
sargan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Snmesp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
SumHes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
summary.plm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
vcovBK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
vcovDC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
vcovG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
vcovHC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
vcovNW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
vcovSCC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Wages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
within_intercept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Index 118
plm-package plm package: linear models for panel data
Description
plm is a package for R which intends to make the estimation of linear panel models straightforward.
plm provides functions to estimate a wide variety of models and to make (robust) inference.
Details
For a gentle and comprehensive introduction to the package, please see the package’s vignette.
The main functions to estimate models are:
plm panel data estimators using lm on transformed data,
pgmm generalized method of moments (GMM) estimation for panel data,
pvcm variable coefficients models for panel data,

4 aneweytest
pmg mean groups (MG), demeaned MG and common correlated effects (CCEMG) estimators.
Next to the model estimation functions, the package offers several functions for statistical tests
related to panel data/models.
Multiple functions for (robust) variance–covariance matrices are at hand as well.
The package also provides data sets to demonstrate functions and to replicate some text book/paper
results. Use data(package="plm") to view a list of available data sets in the package.
Examples
data("Produc", package = "plm")
zz <- plm(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp,
data = Produc, index = c("state","year"))
summary(zz)
# replicates some results from Baltagi (2013), table 3.1

data("Grunfeld", package = "plm")
p <- plm(inv ~ value + capital,
data = Grunfeld, model="pooling")
wi <- plm(inv ~ value + capital,

data = Grunfeld, model="within", effect = "twoways")
swar <- plm(inv ~ value + capital,

data = Grunfeld, model="random", effect = "twoways")
amemiya <- plm(inv ~ value + capital,

data = Grunfeld, model = "random", random.method = "amemiya",
effect = "twoways")
walhus <- plm(inv ~ value + capital,

data = Grunfeld, model = "random", random.method = "walhus",
effect = "twoways")
aneweytest Chamberlain estimator and test for fixed effects
Description
Angrist and Newey’s version of the Chamberlain test
Usage
aneweytest(formula, data, subset, na.action, index = NULL, ...)

Cigar 5
Arguments
formula a symbolic description for the model to be estimated,
data a data.frame,
subset see lm,
na.action see lm,
index the indexes,
... further arguments.
Details
Angrist and Newey’s test is based on the results of the artifactual regression of the within residuals
on the covariates for all the period.
Value
An object of class "htest".
Author(s)
Yves Croissant
References
Angrist, J.D. and Newey, W.K. (1991). Over-identification tests in earnings functions with fixed
effects, Journal of Business & Economic Statistics, 9(3), pp. 317–323.
Examples
data("RiceFarms", package = "plm")
aneweytest(log(goutput) ~ log(seed) + log(totlabor) + log(size), RiceFarms, index = "id")
Cigar Cigarette Consumption
Description
a panel of 46 observations from 1963 to 1992
total number of observations : 1380
observation : regional
country : United States
Usage
data(Cigar)
6 cipstest
Format
A data frame containing :
state state abbreviation

year the year
price price per pack of cigarettes
pop population
pop16 population above the age of 16
cpi consumer price index (1983=100)
ndi per capita disposable income
sales cigarette sales in packs per capita
pimin minimum price in adjoining states per pack of cigarettes
Source
Online complements to Baltagi (2001):
http://www.wiley.com/legacy/wileychi/baltagi/
http://bcs.wiley.com/he-bcs/Books?action=resource&bcsId=4338&itemId=1118672321&
resourceId=13452
References
Baltagi, Badi H. (2001) Econometric Analysis of Panel Data, 2nd ed., John Wiley and Sons.
Baltagi, Badi H. (2013) Econometric Analysis of Panel Data, 5th ed., John Wiley and Sons.
Baltagi, B.H. and D. Levin (1992) “Cigarette taxation: raising revenues and reducing consumption”,
Structural Changes and Economic Dynamics, 3(2), pp. 321–335.
Baltagi, B.H., J.M. Griffin and W. Xiong (2000) “To pool or not to pool: homogeneous versus
heterogeneous estimators applied to cigarette demand”, Review of Economics and Statistics, 82(1),
pp. 117–126.
cipstest Cross-sectionally Augmented IPS Test for Unit Roots in Panel Models
Description
Cross-sectionally augmented Im, Pesaran and Shin (IPS) test for unit roots in panel models.
Usage
cipstest(x, lags = 2, type = c("trend", "drift", "none"),
model = c("cmg", "mg", "dmg"), truncated = FALSE, ...)
cipstest 7
Arguments
x an object of class "pseries",

lags lag order for Dickey-Fuller augmentation,
type one of "trend", "drift", "none",
model one of "cmg", "mg", "dmg",
truncated logical specifying whether to calculate the truncated version of the test,
... further arguments passed to critvals.
Details
This cross-sectionally augmented version of the IPS unit root test (H0: the pseries has a unit
root) is a so-called second-generation panel unit root test: it is in fact robust against cross-sectional
dependence, provided that the default type="cmg" is calculated. Else one can obtain the standard
(model="mg") or cross-sectionally demeaned (model="dmg") versions of the IPS test.
Value
Author(s)
Giovanni Millo
References
Pesaran, M.H. (2007) A simple panel unit root test in the presence of cross-section dependence,
Journal of Applied Econometrics, 22(2), pp. 265–312.
See Also
purtest
Examples

Produc <- pdata.frame(Produc, index=c("state", "year"))
## check whether the gross state product (gsp) is trend-stationary
cipstest(Produc$gsp, type = "trend")
8 Crime
Crime Crime in North Carolina
Description
a panel of 90 observational units (counties) from 1981 to 1987
Usage
data(Crime)
Format
county county identifier
year year from 1981 to 1987
crmrte crimes committed per person
prbarr ’probability’ of arrest
prbconv ’probability’ of conviction
prbpris ’probability’ of prison sentence
avgsen average sentence, days
polpc police per capita
density people per square mile
taxpc tax revenue per capita
region factor. One of ’other’, ’west’ or ’central’.
smsa factor. Does the individual reside in a SMSA (standard metropolitan statistical area)?
pctmin percentage minority in 1980
wcon weekly wage in construction
wtuc weekly wage in trns, util, commun
wtrd weekly wage in whole sales and retail trade
wfir weekly wage in finance, insurance and real estate
wser weekly wage in service industry
wmfg weekly wage in manufacturing
wfed weekly wage of federal employees
wsta weekly wage of state employees
wloc weekly wage of local governments employees
mix offence mix: face-to-face/other
pctymle percentage of young males
detect_lin_dep 9
Source
resourceId=13452
See also Journal of Applied Econometrics data archive entry for Baltagi (2006) at http://qed.
econ.queensu.ca/jae/2006-v21.4/baltagi/.
References
Cornwell, C. and W.N. Trumbull (1994) “Estimating the economic model of crime with panel data”,
Review of Economics and Statistics, 76(2), pp. 360–366.
Baltagi, B. H. (2006) “Estimating an economic model of crime using panel data from North Car-
olina”, Journal of Applied Econometrics, 21(4), pp. 543–547.
detect_lin_dep Functions to detect linear dependence
Description
Little helper functions to aid users to detect linear dependent columns in a two-dimensional data
structure, especially in a (transformed) model matrix - typically useful in interactive mode during
model building phase.
Usage
detect_lin_dep(object, ...)
## S3 method for class 'matrix'
detect_lin_dep(object, suppressPrint = FALSE, ...)
## S3 method for class 'data.frame'
## S3 method for class 'plm'

alias(object, ...)
## S3 method for class 'pFormula'
alias(object, data, model = c("pooling", "within", "Between", "between",
"mean", "random", "fd"),
effect = c("individual", "time", "twoways"), ...)
10 detect_lin_dep
Arguments
object for detect_lin_dep: an object which should be checked for linear dependence
(of class "matrix", "data.frame", or "plm"); for alias: either an estimated
model of class "plm" or a "pFormula". Usually, one wants to input a model
matrix here or check an already estimated plm model,
suppressPrint for detect_lin_dep only: logical indicating whether a message shall be printed;
defaults to printing the message, i. e. to suppressPrint = FALSE,
data for alias.pFormula only: model frame (created by plm’s model.frame method),
model, effect for alias.pFormula only: model and effect to specify the data transformation,
see plm,
Details
Linear dependence of columns/variables is (usually) readily avoided when building one’s model.
However, linear dependence is sometimes not obvious and harder to detect for less experienced
applied statisticians. The so called "dummy variable trap" is a common and probably the best–
known fallacy of this kind (see e. g. Wooldridge (2016), sec. 7-2.). When building linear models
with lm or plm’s pooling model, linear dependence in one’s model is easily detected, at times post
hoc.
However, linear dependence might also occur after some transformations of the data, albeit it is not
present in the untransformed data. The within transformation (also called fixed effect transforma-
tion) used in the "within" model can result in such linear dependence and this is harder to come to
mind when building a model. See Examples for two examples of linear dependent columns after
the within transformation: ex. 1) the transformed variables have the opposite sign of one another;
ex. 2) the transformed variables are identical.
During plm’s model estimation, linear dependent columns and their corresponding coefficients in
the resulting object are silently dropped, while the corresponding model frame and model matrix
still contain the affected columns. The plm object contains an element aliased which indicates
any such aliased coefficients by a named logical.
Both functions, detect_lin_dep and alias, help to detect linear dependence and accomplish al-
most the same: detect_lin_dep is a stand alone implementation while alias is a wrapper around
alias.lm, extending the alias generic to classes "plm" and "pFormula". alias hinges on the
availability of the package MASS on the system. Not all arguments of alias.lm are supported.
Output of alias is more informative as it gives the linear combination of dependent columns (after
data transformations, i. e. after (quasi)-demeaning) while detect_lin_dep only gives columns
involved in the linear dependence in a simple format (thus being more suited for automatic post–
processing of the information).
Value
For detect_lin_dep: A named numeric vector containing column numbers of the linear dependent
columns in the object after data transformation, if any are present. NULL if no linear dependent
columns are detected.
For alias: return value of alias.lm run on the (quasi-)demeaned model, i. e. the information
outputted applies to the transformed model matrix, not the original data.
detect_lin_dep 11
Author(s)
Kevin Tappe
References
Wooldridge, J.M. (2016) Introductory Econometrics: A Modern Approach, 6th ed., Cengage Learn-
ing, Boston, sec. 7-2, pp. 206–211.
See Also
alias. model.matrix and especially plm’s model.matrix for (transformed) model matrices.
model.frame.
Examples
### Example 1 ###
# prepare the data
data(Cigar)
Cigar[ , "fact1"] <- c(0,1)
Cigar[ , "fact2"] <- c(1,0)
Cigar.p <- pdata.frame(Cigar)
# setup a pFormula and a model frame

pform <- pFormula(price ~ 0 + cpi + fact1 + fact2)
mf <- model.frame(pform, data = Cigar.p)
# no linear dependence in the pooling model's model matrix

# (with intercept in the formula, there would be linear depedence)
detect_lin_dep(model.matrix(pform, data = mf, model = "pooling"))
# linear dependence present in the FE transformed model matrix

modmat_FE <- model.matrix(pform, data = mf, model = "within")
detect_lin_dep(modmat_FE)
mod_FE <- plm(pform, data = Cigar.p, model = "within")
detect_lin_dep(mod_FE)
alias(mod_FE) # => fact1 == -1*fact2
plm(pform, data = mf, model = "within")$aliased # "fact2" indicated as aliased
# look at the data: after FE transformation fact1 == -1*fact2

head(modmat_FE)
all.equal(modmat_FE[ , "fact1"], -1*modmat_FE[ , "fact2"])
### Example 2 ###

# Setup the data:
# Assume CEOs stay with the firms of the Grunfeld data
# for the firm's entire lifetime and assume some fictional
# data about CEO tenure and age in year 1935 (first observation
# in the data set) to be at 1 to 10 years and 38 to 55 years, respectively.
# => CEO tenure and CEO age increase by same value (+1 year per year).
data(Grunfeld, package = "plm")
set.seed(42)
# add fictional data
12 dynformula
Grunfeld$CEOtenure <- c(replicate(10, seq(from=s<-sample(1:10, 1), to=s+19, by=1)))

Grunfeld$CEOage <- c(replicate(10, seq(from=s<-sample(38:65, 1), to=s+19, by=1)))
# look at the data

head(Grunfeld, 50)
pform <- pFormula(inv ~ value + capital + CEOtenure + CEOage)

mf <- model.frame(pform, data=pdata.frame(Grunfeld))
# no linear dependent columns in original data/pooling model

modmat_pool <- model.matrix(pform, data = mf, model="pooling")
detect_lin_dep(modmat_pool)
mod_pool <- plm(pform, data = Grunfeld, model = "pooling")
alias(mod_pool)
# CEOtenure and CEOage are linear dependent after FE transformation

# (demeaning per individual)
modmat_FE <- model.matrix(pform, data = mf, model="within")
detect_lin_dep(modmat_FE)
mod_FE <- plm(pform, data = Grunfeld, model = "within")
detect_lin_dep(mod_FE)
alias(mod_FE)
# look at the transformed data: after FE transformation CEOtenure == 1*CEOage

head(modmat_FE, 50)
all.equal(modmat_FE[ , "CEOtenure"], modmat_FE[ , "CEOage"])
dynformula Dynamic Formula
Description
A function to easily create a formula with lags and differences
Usage
dynformula(formula, lag.form = NULL, diff.form = NULL, log.form = NULL)
Arguments
formula a formula,
lag.form a list containing the lag structure of each variable in the formula,
diff.form a vector (or a list) of logical values indicating whether variables should be dif-
ferenced,
log.form a vector (or a list) of logical values indicating whether variables should be in
logarithms.
EmplUK 13
Details
lag.form is a list, diff.form and log.form are vectors (or lists) that should be of length equal to
the total number of variables. Each element of these lists/vectors is:
• either a vector of length 2 (c(1,4) means lags 1,2,3 and 4) or a scalar (3 means lags 0,1,2,3
except for the left–hand side variable for which it is 1,2,3) for lag.form.
• logical values for diff.form and log.form.
It can also be an incomplete named list/vector (for example, to apply the transformation for only
some variables) with eventually an unnamed element which then is the default value.
Value
An object of class c("dynformula","formula"), which is a formula with four additional at-
tributes: var, the names of the variables in the formula, lag, diff, and log, which store the in-
formation about lags, differences and logs, respectively.
A formula method coerces the dynformula object to a standard formula.
Author(s)
Yves Croissant
Examples
# all variables in log, x1, x2 and x3 laged twice, y laged once and x3 differenced
z <- dynformula(y ~ x1 + x2 + x3, lag.form = list(2, y = 1),
diff.form = c(x3 = TRUE), log.form = TRUE)
formula(z)
EmplUK Employment and Wages in the United Kingdom
Description
An unbalanced panel of 140 observations from 1976 to 1984
observation : firms
country : United Kingdom
Usage
data(EmplUK)
14 ercomp
Format
firm firm index
year year
sector the sector of activity
emp employment
wage wages
capital capital
output output
Source
Arellano, M. and Bond, S. (1991), Some Tests of Specification for Panel Data: Monte Carlo Evi-
dence and an Application to Employment Equations, The Review of Economic Studies, vol. 58(2),
1991, pp. 227–297.
ercomp Estimation of the error components
Description
This function enables the estimation of the variance components of a panel model.
Usage
ercomp(object, ...)
## S3 method for class 'formula'
ercomp(object, data, effect = c("individual", "time", "twoways"),
method = c("swar", "walhus", "amemiya","nerlove"),
dfcor = NULL, index = NULL, ...)
ercomp(object, ...)
## S3 method for class 'ercomp'
print(x, digits= max(3, getOption("digits") - 3), ...)
Arguments
object a formula or a plm object,
data a data.frame,
effect the effects introduced in the model, see lm for details,
method method of estimation for the variance components, see lm for details,
dfcor a numeric vector of length 2 indicating which degree of freedom should be used,
index the indexes,
ercomp 15
x a ercomp object,
digits digits,
Value
An object of class "ercomp": a list containing a list called sigma2 which contains the estimates
of the variance components, and theta which is the parameters used for the transformation of the
variables.
Author(s)
Yves Croissant
References
Amemiya, T. (1971) The estimation of the variances in a variance–components model, International

Economic Review, 12(1), pp. 1–13.
Nerlove, M. (1971) Further evidence on the estimation of dynamic economic relations from a time–
series of cross–sections, Econometrica, 39(2), pp. 359–382.
Swamy, P.A.V.B. and Arora, S.S. (1972) The exact finite sample properties of the estimators of
coefficients in the error components regression models, Econometrica, 40(2), pp. 261–275.
Wallace, T.D. and Hussain, A. (1969) The use of error components models in combining cross
section with time series data, Econometrica, 37(1), pp. 55–72.
See Also
plm where the estimates of the variance components are used if a random effects model is estimated
Examples
# an example of the formula method
ercomp(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp, data = Produc,
method = "walhus", effect = "time")
# same with the plm method
z <- plm(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp,
data = Produc, random.method = "walhus",
effect = "time", model = "random")
ercomp(z)
# a two-ways model
ercomp(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp, data = Produc,
method = "amemiya",effect = "twoways")
16 fixef.plm
fixef.plm Extract the Fixed Effects
Description
Function to extract the fixed effects from a plm object and associated summary method.
Usage
fixef(object, effect = NULL,
type = c("level", "dfirst", "dmean"), vcov = NULL, ...)
## S3 method for class 'fixef'
print(x, digits = max(3, getOption("digits") - 2),
width = getOption("width"), ...)
## S3 method for class 'fixef'
summary(object, ...)
## S3 method for class 'summary.fixef'
print(x, digits = max(3, getOption("digits") -
2), width = getOption("width"), ...)
Arguments
x,object an object of class "plm", an object of class "fixef" for the print and the
summary method,
effect one of "individual" or "time", only relevant in case of two–ways effects mod-
els,
vcov a variance–covariance matrix furnished by the user or a function to calculate one
(see Examples),
type one of "level", "dfirst", or "dmean",
digits digits,
width the maximum length of the lines in the print output,
Details
Function fixef calculates the fixed effects and returns an object of class c("fixef", "numeric").
With the type argument, the fixed effects may be returned in levels ("level"), as deviations from
the first value of the index ("dfirst"), or as deviations from the overall mean ("dmean"). If the
argument vcov was specified, the standard errors (stored as attribute "se" in the return value) are the
respective robust standard errors.
The associated summary method returns an extended object of class c("summary.fixef", "matrix")
with more information (see sections Value and Examples).
References with formulae (except for the two-ways unbalanced case) are, e.g., Greene (2012), Ch.
11.4.4, p. 364, formulae (11-25); Wooldridge (2010), Ch. 10.5.3, pp. 308-309, formula (10.58).
fixef.plm 17
Value
For function fixef an object of class c("fixef", "numeric") is returned:
It is a numeric vector containing the fixed effects with attribute se which contains the standard er-
rors. There are two further attributes: attribute type contains the chosen type (the value of argument
type as a character); attribute df.residual holds the residual degrees of freedom (integer) from
the fixed effects model (plm object) on which fixef was run.
For function summary.fixef an object of class c("summary.fixef", "matrix" is returned:
It is a matrix with four columns in this order: the estimated fixed effects, their standard errors
and associated t–values and p–values. The type of the fixed effects and the standard errors in the
summary.fixef objects corresponds to was requested in the fixef function by arguments type and
vcov.
Author(s)
Yves Croissant
References
Greene, W.H. (2012), Econometric Analysis, 7th ed., Prentice Hall, Ch. 11.4.4.
Wooldridge, J.M. (2010) Econometric Analysis of Cross-Section and Panel Data, 2nd ed., MIT
Press, Ch. 10.5.3.
See Also
within_intercept for the overall intercept of fixed effect models along its standard error, plm for
plm objects and within models (= fixed effects models) in general.
Examples
gi <- plm(inv ~ value + capital, data = Grunfeld, model = "within")
fixef(gi)
summary(fixef(gi))
summary(fixef(gi))[ , c("Estimate", "Pr(>|t|)")] # only estimates and p-values
# relationship of type = "dmean" and "level" and overall intercept

fx_level <- fixef(gi, type = "level")
fx_dmean <- fixef(gi, type = "dmean")
overallint <- within_intercept(gi)
all.equal(overallint + fx_dmean, fx_level, check.attributes = FALSE) # TRUE
# extract time effects in a twoways effects model

gi_tw <- plm(inv ~ value + capital, data = Grunfeld,
model = "within", effect = "twoways")
fixef(gi_tw, effect = "time")
# with supplied variance-covariance matrix as matrix, function,

# and function with additional arguments
fx_level_robust1 <- fixef(gi, vcov = vcovHC(gi))
fx_level_robust2 <- fixef(gi, vcov = vcovHC)
18 Gasoline
fx_level_robust3 <- fixef(gi, vcov = function(x) vcovHC(x, method = "white2"))

summary(fx_level_robust1) # gives fixed effects, robust SEs, t- and p-values
# calc. fitted values of oneway within model:

fixefs <- fixef(gi)[index(gi, which = "id")]
fitted_by_hand <- fixefs + gi$coefficients["value"] * gi$model$value +
gi$coefficients["capital"] * gi$model$capital
Gasoline Gasoline Consumption
Description
A panel of 18 observations from 1960 to 1978
observation : country
country : OECD
Usage
data(Gasoline)
Format
country a factor with 18 levels
year the year
lgaspcar logarithm of motor gasoline consumption per car
lincomep logarithm of real per-capita income
lrpmg logarithm of real motor gasoline price
lcarpcap logarithm of the stock of cars per capita
Source
resourceId=13452
References
Baltagi, B.H. and J.M. Griffin (1983) “Gasoline demand in the OECD: An application of pooling
and testing procedures”, European Economic Review, 22(2), pp. 117–137.
Grunfeld 19
Grunfeld Grunfeld’s Investment Data
Description
A balanced panel of 10 observational units (firms) from 1935 to 1954
observation : production units
Usage
data(Grunfeld)
Format
firm observation
year date
inv gross Investment
value value of the firm
capital stock of plant and equipment
Note
The Grunfeld data as provided in package plm is the same data as used in Baltagi (2001), see
Examples below.
NB:
Various versions of the Grunfeld data circulate online. Also, various text books (and also varying
among editions) and papers use different subsets of the original Grunfeld data, some of which
contain errors in a few data points compared to the original data used by Grunfeld (1958) in his
PhD thesis. See Kleiber/Zeileis (2010) and its accompanying website for a comparison of various
Grunfeld data sets in use.
Source
http://www.wiley.com/legacy/wileychi/baltagi/ http://www.wiley.com/legacy/wileychi/
baltagi/supp/Grunfeld.fil
resourceId=13452
20 Hedonic
References
Baltagi, Badi H. (2001) Econometric Analysis of Panel Data, 2nd ed., John Wiley and Sons;
accompanying website http://www.wiley.com/legacy/wileychi/baltagi/.
Baltagi, Badi H. (2013) Econometric Analysis of Panel Data, 5th ed., John Wiley and Sons;
accompanying website http://bcs.wiley.com/he-bcs/Books?action=resource&bcsId=4338&
itemId=1118672321&resourceId=13452.
Grunfeld, Yehuda (1958) The Determinants of Corporate Investment, Ph.D. thesis, Department of
Economics, University of Chicago.
Kleiber, C./Zeileis, A. (2010) “The Grunfeld Data at 50”, German Economic Review, 11(4), pp.
404–417, http://dx.doi.org/10.1111/j.1468-0475.2010.00513.x;
website accompanying the paper with various variants of the Grunfeld data: http://statmath.
wu-wien.ac.at/~zeileis/grunfeld/.
See Also
For the complete Grunfeld data (11 firms), see Grunfeld, in the AER package.
Examples
## Not run:
# Compare plm's Grunfeld data to Baltagi's (2001) Grunfeld data:
data(Grunfeld, package="plm")
Grunfeld_baltagi2001 <- read.csv("http://www.wiley.com/legacy/wileychi/
baltagi/supp/Grunfeld.fil", sep="", header = FALSE)
library(compare)
compare::compare(Grunfeld, Grunfeld_baltagi2001, allowAll = T) # same data set
## End(Not run)
Hedonic Hedonic Prices of Census Tracts in the Boston Area
Description
A cross-section
number of observations : 506
Usage
data(Hedonic)
Hedonic 21
Format
A dataframe containing:
mv median value of owner–occupied homes

crim crime rate
zn proportion of 25,000 square feet residential lots
indus proportion of no–retail business acres
chas is the tract bounds the Charles River?
nox annual average nitrogen oxide concentration in parts per hundred million
rm average number of rooms
age proportion of owner units built prior to 1940
dis weighted distances to five employment centers in the Boston area
rad index of accessibility to radial highways
tax full value property tax rate ($/$10,000)
ptratio pupil/teacher ratio
blacks proportion of blacks in the population
lstat proportion of population that is lower status
townid town identifier
Source

resourceId=13452
References
Belsley, D.A., E. Kuh and R. E. Welsch (1980): Regression Diagnostics: Identifying Influential
Data and Sources of Collinearity, John Wiley, New York.
Harrison, D. and D.L. Rubinfeld (1978): “Hedonic housing prices and the demand for clean air”,
Journal of Environmental Economics and Management, 5(1), pp. 81–102.
22 index
index Extract the indexes of panel data
Description
This function extracts the information about the structure of the individual and time dimensions of
panel data.
Usage
## S3 method for class 'pindex'
index(x, which = NULL, ...)
## S3 method for class 'pdata.frame'
## S3 method for class 'pseries'
## S3 method for class 'panelmodel'
Arguments
x an object of class "pindex", "pdata.frame", "pseries" or "panelmodel",
which the index(es) to be extracted (see details),
Details
panel data are stored in a "pdata.frame" which has an "index" attribute. Fitted models in "plm"
has a "model" element which is also a "pdata.frame" and therefore also have an "index" at-
tribute. Finally, each series in a "pdata.frame" is of class "pseries", which also has this "index"
attribute. "index" methods are available for all these objects. The argument "which" indicates
which index should be extracted. If which = NULL, both indexes are extracted, but "which" can
also be a vector of length 1 or 2 containing either characters (the names of the individual variable
and/or of the time variable or "id" and "time") or integers (1 for the individual index and 2 for the
time index.)
Value
a vector or a "data.frame" containing either one index or both indexes.
Author(s)
Yves Croissant
See Also
plm
is.pbalanced 23
Examples
Gr <- pdata.frame(Grunfeld, index = c("firm", "year"))
m <- plm(inv ~ value + capital, data = Gr)
index(Gr, "firm")
index(Gr, "time")
index(Gr$inv, c(2, 1))
index(m, "id")
is.pbalanced Check if data are balanced
Description
This function checks if the data are balanced, i.e. each individual has the same time periods
Usage
is.pbalanced(x, ...)
is.pbalanced(x, index = NULL, ...)
## S3 method for class 'pgmm'
Arguments
x an object of class pdata.frame, data.frame, pseries, panelmodel, pgmm;
index only relevant for data.frame interface; if NULL, the first two columns of the
data.frame are assumed to be the index variables; if not NULL, both dimensions
(’individual’, ’time’) need to be specified by index as character of length 2 for
data frames, for further details see pdata.frame,
Details
Balanced data are data for which each individual has the same time periods. The returned values of
the is.pbalanced(object) methods are identical to pdim(object)$balanced. is.pbalanced is
provided as a short cut and is faster than pdim(object)$balanced because it avoids those compu-
tations performed by pdim which are unnecessary to determine the balancedness of the data.
Value
A logical indicating whether the data associated with object x are balanced (TRUE) or not (FALSE).
24 is.pconsecutive
See Also
punbalancedness for two measures of unbalancedness, make.pbalanced to make data balanced;

is.pconsecutive to check if data are consecutive; make.pconsecutive to make data consecutive
(and, optionally, also balanced).
pdim to check the dimensions of a ’pdata.frame’ (and other objects), pvar to check for individual
and time variation of a ’pdata.frame’ (and other objects), pseries, data.frame, pdata.frame.
Examples
# take balanced data and make it unbalanced

# by deletion of 2nd row (2nd time period for first individual)
Grunfeld_missing_period <- Grunfeld[-2, ]
is.pbalanced(Grunfeld_missing_period) # check if balanced: FALSE
pdim(Grunfeld_missing_period)$balanced # same
# pdata.frame interface
pGrunfeld_missing_period <- pdata.frame(Grunfeld_missing_period)
is.pbalanced(Grunfeld_missing_period)
# pseries interface
is.pbalanced(pGrunfeld_missing_period$inv)
is.pconsecutive Check if time periods are consecutive
Description
This function checks for each individual if its associated time periods are consecutive (no "gaps" in
time dimension per individual)
Usage

is.pconsecutive(x, na.rm.tindex = FALSE, ...)
is.pconsecutive(x, index = NULL, na.rm.tindex = FALSE, ...)
## Default S3 method:
is.pconsecutive(x, id, time, na.rm.tindex = FALSE, ...)
is.pconsecutive 25
Arguments
x usually, an object of class pdata.frame, data.frame, pseries, or an estimated
panelmodel; for the default method x can also be an arbitrary vector or NULL,
see Details,
na.rm.tindex logical indicating whether any NA values in the time index are removed before
consecutiveness is evaluated (defaults to FALSE),
(’individual’, ’time’) need to be specified by index for is.pconsecutive on
id, time only relevant for default method: vectors specifying the id and time dimensions,
i. e. a sequence of individual and time identifiers, each as stacked time series,
Details
(p)data.frame, pseries and estimated panelmodel objects can be tested if their time periods are con-
secutive per individual. For evaluation of consecutiveness, the time dimension is interpreted to be
numeric, and the data are tested for being a regularly spaced sequence with distance 1 between
the time periods for each individual (for each individual the time dimension can be interpreted as
sequence t, t+1, t+2, . . . where t is an integer). As such, the "numerical content" of the time index
variable is considered for consecutiveness, not the "physical position" of the various observations
for an individuals in the (p)data.frame/pseries (it is not about "neighbouring" rows).
The default method also works for argument x being an arbitrary vector (see Examples), provided
one can supply arguments id and time, which need to ordered as stacked time series. As only id
and time are really necessary for the default method to evaluate the consecutiveness, x = NULL is
also possible. However, if the vector x is also supplied, additional input checking for equality of the
lengths of x, id and time is performed, which is safer.
For the data.frame interface, the data is ordered in the appropriate way (stacked time series) be-
fore the consecutiveness is evaluated. For the pdata.frame and pseries interface, ordering is not
performed because both data types are already ordered in the appropriate way when created.
Note: Only the presence of the time period itself in the object is tested, not if there are any other
variables. NA values in individual index are not examined but silently dropped - In this case, it is
not clear which individual is meant by id value NA, thus no statement about consecutiveness of time
periods for those "NA-individuals" is possible.
Value
A named logical vector (names are those of the individuals). The i-th element of the returned
vector corresponds to the i-th individual. The values of the i-th element can be:
TRUE if the i-th individual has consecutive time periods,

FALSE if the i-th individual has non-consecutive time periods,
NA if there are any NA values in time index of the i-th the individual; see also
argument na.rm.tindex to remove those.
26 is.pconsecutive
Author(s)
Kevin Tappe
See Also
make.pconsecutive to make data consecutive (and, as an option, balanced at the same time) and
make.pbalanced to make data balanced.
pdim to check the dimensions of a ’pdata.frame’ (and other objects), pvar to check for individual
and time variation of a ’pdata.frame’ (and other objects), lag for lagged (and leading) values of a
’pseries’ object.
pseries, data.frame, pdata.frame, for class ’panelmodel’ see plm and pgmm.
Examples
is.pconsecutive(Grunfeld)
is.pconsecutive(Grunfeld, index=c("firm", "year"))
# delete 2nd row (2nd time period for first individual)

# -> non consecutive
is.pconsecutive(Grunfeld_missing_period)
all(is.pconsecutive(Grunfeld_missing_period)) # FALSE
# delete rows 1 and 2 (1st and 2nd time period for first individual)
# -> consecutive
Grunfeld_missing_period_other <- Grunfeld[-c(1,2), ]
is.pconsecutive(Grunfeld_missing_period_other) # all TRUE
# delete year 1937 (3rd period) for _all_ individuals

Grunfeld_wo_1937 <- Grunfeld[Grunfeld$year != 1937, ]
is.pconsecutive(Grunfeld_wo_1937) # all FALSE
pGrunfeld <- pdata.frame(Grunfeld)
is.pconsecutive(pGrunfeld) # all TRUE
is.pconsecutive(pGrunfeld_missing_period) # first FALSE, others TRUE
# panelmodel interface (first, estimate some models)

mod_pGrunfeld <- plm(inv ~ value + capital, data = Grunfeld)
mod_pGrunfeld_missing_period <- plm(inv ~ value + capital, data = Grunfeld_missing_period)
is.pconsecutive(mod_pGrunfeld)
is.pconsecutive(mod_pGrunfeld_missing_period)
nobs(mod_pGrunfeld) # 200
nobs(mod_pGrunfeld_missing_period) # 199
LaborSupply 27
# pseries interface
pinv <- pGrunfeld$inv
pinv_missing_period <- pGrunfeld_missing_period$inv
is.pconsecutive(pinv)
is.pconsecutive(pinv_missing_period)
# default method for arbitrary vectors or NULL

inv <- Grunfeld$inv
inv_missing_period <- Grunfeld_missing_period$inv
is.pconsecutive(inv, id = Grunfeld$firm, time = Grunfeld$year)
is.pconsecutive(inv_missing_period, id = Grunfeld_missing_period$firm,
time = Grunfeld_missing_period$year)
# (not run) demonstrate mismatch lengths of x, id, time

# is.pconsecutive(x = inv_missing_period, id = Grunfeld$firm, time = Grunfeld$year)
# only id and time are needed for evaluation

is.pconsecutive(NULL, id = Grunfeld$firm, time = Grunfeld$year)
LaborSupply Wages and Hours Worked
Description
Usage
data(LaborSupply)
Format
lnhr log of annual hours worked

lnwg log of hourly wage
kids number of children
age age
disab bad health
id id
year year
28 lag
Source
Online complements to Ziliak (1997).
Journal of Business Economics and Statistics web site: http://amstat.tandfonline.com/loi/
ubes20/.
References
Cameron, A.C. and P.K. Trivedi (2005) Microeconometrics : methods and applications, Cambridge,
pp. 708–15, 754–6.
Ziliak, J.P. (1997) “Efficient Estimation With Panel Data when Instruments are Predetermined:
An Empirical Comparison of Moment-Condition Estimators”, Journal of Business and Economic
Statistics, 15(4), pp. 419–431.
lag lag, lead, and diff for panel data
Description
lag, lead, and diff functions for class pseries.
Usage
lag(x, k = 1, ...)
lead(x, k = 1, ...)
diff(x, lag = 1, ...)
Arguments
x a pseries object,
k an integer vector, the number of lags for the lag and lead methods (can also
be negative). For the lag method, a positive (negative) k gives lagged (leading)
values. For the lead method, a positive (negative) k gives leading (lagged)
values, thus, lag(x, k = -1) yields the same as lead(x, k = 1). If k is an
integer vector with length > 1 (k = c(k1, k2, ...)) a matrix with multiple
lagged pseries is returned,
lag the number of lags for the diff method (only non–negative values for parameter
lag are allowed in diff),
Value
These functions return an object of class pseries except if lag (thus also lead) is called with more
than one lag (lead), i. e. length(k) > 1, a matrix is returned.
make.pbalanced 29
Note
The sign of k in lag.pseries results in inverse behavior compared to lag and lag.zoo.
Author(s)
Yves Croissant
See Also
For further function for ’pseries’ objects: between, Between, Within, summary.pseries, print.summary.pseries,
as.matrix.pseries. To check if the time periods are consecutive per individual, see is.pconsecutive.
Examples
# First, create a pdata.frame
data("EmplUK", package = "plm")
Em <- pdata.frame(EmplUK)
# Then extract a series, which becomes additionally a pseries

z <- Em$output
class(z)
# compute the first and third lag, and the difference lagged twice
lag(z)
lag(z, 3)
diff(z, 2)
# compute negative lags (= leading values)

lag(z, -1)
lead(z, 1) # same as line above
identical(lead(z, 1), lag(z, -1)) # TRUE
# compute more than one lag at once

lag(z, c(1,2))
make.pbalanced Make data balanced
Description
This function makes the data balanced, i.e. each individual has the same time periods
Usage
make.pbalanced(x, balance.type = c("fill", "shared"), ...)
make.pbalanced(x, balance.type = c("fill", "shared"), ...)
make.pbalanced(x, balance.type = c("fill", "shared"), index = NULL, ...)
30 make.pbalanced
Arguments
x an object of class pdata.frame, data.frame, or pseries;

balance.type character, one of "fill", "shared", see Detail,
Details
(p)data.frame and pseries objects are made balanced, meaning each individual has the same time
periods. Depending on the value of balance.type, the balancing is done in two different ways:
• balance.type = "fill" (default): The union of available time periods over all individuals
is taken (w/o NA values). Missing time periods for an individual are identified and correspond-
ing rows (elements for pseries) are inserted and filled with NA for the non–index variables
(elements for a pseries). This means, only time periods present for at least one individual are
inserted, if missing.
• balance.type = "shared": The intersect of available time periods over all individuals is
taken (w/o NA values). Thus, time periods not available for all individuals are discarded (only
time periods shared among all individuals are left in the result).
The data are not necessarily made consecutive (regularly time series with distance 1), because bal-
ancedness does not imply consecutiveness. For making the data consecutive, use make.pconsecutive
(and, optionally, set argument balanced = TRUE to make consecutive and balanced, see also Ex-
amples for a comparison of the two functions.
Note: rows of (p)data.frames (elements for pseries) with NA values in individual or time index are not
examined but silently dropped before the data are made balanced. In this case, it cannot be inferred
which individual or time period is meant by the missing value(s) (see also Examples). Especially,
this means: NA values in the first/last position of the original time periods for an individual are
dropped, which are usually meant to depict the beginning and ending of the time series for that
individual. Thus, one might want to check if there are any NA values in the index variables before
applying make.pbalanced, and especially check for NA values in the first and last position for each
individual in original data and, if so, maybe set those to some meaningful begin/end value for the
time series.
Value
An object of the same class as the input x, i.e. a pdata.frame, data.frame or a pseries which is made
balanced based on the index variables. The returned data are sorted as a stacked time series.
Author(s)
Kevin Tappe
make.pbalanced 31
See Also
is.pbalanced to check if data are balanced; is.pconsecutive to check if data are consecutive;
make.pconsecutive to make data consecutive (and, optionally, also balanced).
punbalancedness for two measures of unbalancedness, pdim to check the dimensions of a ’pdata.frame’
(and other objects), pvar to check for individual and time variation of a ’pdata.frame’ (and other
objects), lag for lagged (and leading) values of a ’pseries’ object.
pseries, data.frame, pdata.frame.
Examples
# take data and make it unbalanced
nrow(Grunfeld) # 200 rows
pdim(Grunfeld_missing_period)$balanced # check if balanced: FALSE
make.pbalanced(Grunfeld_missing_period) # make it balanced (by filling)
make.pbalanced(Grunfeld_missing_period, balance.type = "shared") # (shared periods)
nrow(make.pbalanced(Grunfeld_missing_period))
nrow(make.pbalanced(Grunfeld_missing_period, balance.type = "shared"))
# more complex data:

# First, make data unbalanced (and non-consecutive)
# by deletion of 2nd time period (year 1936) for all individuals
# and more time periods for first individual only
Grunfeld_unbalanced <- Grunfeld[Grunfeld$year != 1936, ]
Grunfeld_unbalanced <- Grunfeld_unbalanced[-c(1,4), ]
pdim(Grunfeld_unbalanced)$balanced # FALSE
all(is.pconsecutive(Grunfeld_unbalanced)) # FALSE
g_bal <- make.pbalanced(Grunfeld_unbalanced)

pdim(g_bal)$balanced # TRUE
unique(g_bal$year) # all years but 1936
nrow(g_bal) # 190 rows
head(g_bal) # 1st individual: years 1935, 1939 are NA
# NA in 1st, 3rd time period (years 1935, 1937) for first individual
Grunfeld_NA <- Grunfeld
Grunfeld_NA[c(1, 3), "year"] <- NA
g_bal_NA <- make.pbalanced(Grunfeld_NA)
head(g_bal_NA) # years 1935, 1937: NA for non-index vars
nrow(g_bal_NA) # 200
make.pbalanced(Grunfeld_missing_period)
# pseries interface
make.pbalanced(pGrunfeld_missing_period$inv)
# comparison to make.pconsecutive
g_consec <- make.pconsecutive(Grunfeld_unbalanced)
32 make.pconsecutive
all(is.pconsecutive(g_consec)) # TRUE
pdim(g_consec)$balanced # FALSE
head(g_consec, 22) # 1st individual: no years 1935/6; 1939 is NA;
# other indviduals: years 1935-1954, 1936 is NA
nrow(g_consec) # 198 rows
g_consec_bal <- make.pconsecutive(Grunfeld_unbalanced, balanced = TRUE)

all(is.pconsecutive(g_consec_bal)) # TRUE
pdim(g_consec_bal)$balanced # TRUE
head(g_consec_bal) # year 1936 is NA for all individuals
nrow(g_consec_bal) # 200 rows
head(g_bal) # no year 1936 at all

make.pconsecutive Make data consecutive (and, optionally, also balanced)
Description
This function makes the data consecutive for each individual (no "gaps" in time dimension per
individual) and, optionally, also balanced
Usage

make.pconsecutive(x, balanced = FALSE, ...)
make.pconsecutive(x, balanced = FALSE, ...)
make.pconsecutive(x, balanced = FALSE, index = NULL, ...)
Arguments
x an object of class pdata.frame, data.frame, or pseries,

balanced logical, indicating whether the data should _additionally_ be made balanced
(default: FALSE),
make.pconsecutive 33
Details
(p)data.frame and pseries objects are made consecutive, meaning their time periods are made con-
secutive per individual. For consecutiveness, the time dimension is interpreted to be numeric,
and the data are extended to a regularly spaced sequence with distance 1 between the time peri-
ods for each individual (for each individual the time dimension become a sequence t, t+1, t+2,
. . . where t is an integer). Non–index variables are filled with NA for the inserted elements (rows for
(p)data.frames, vector elements for pseries).
With argument balanced = TRUE, additionally to be made consecutive, the data also can be made
a balanced panel/pseries. Note: This means consecutive AND balanced; balancedness does not
imply consecutiveness. In the result, each individual will have the same time periods in their
time dimension by taking the min and max of the time index variable over all individuals (w/o
NA values) and inserting the missing time periods. Looking at the number of rows of the resulting
(pdata.frame) (elements for pseries), this results in nrow(make.pconsecutive, balanced = FALSE)
<= nrow(make.pconsecutive, balanced = TRUE). For making the data only balanced, i.e. not de-
manding consecutiveness at the same time, use make.pbalanced (see Examples for a compari-
son)).
Note: rows of (p)data.frames (elements for pseries) with NA values in individual or time index are
not examined but silently dropped before the data are made consecutive. In this case, it is not clear
which individual or time period is meant by the missing value(s). Especially, this means: If there
are NA values in the first/last position of the original time periods for an individual, which usually
depicts the beginning and ending of the time series for that individual, the beginning/end of the
resulting time series is taken to be the min and max (w/o NA values) of the original time series for
that individual, see also Examples. Thus, one might want to check if there are any NA values in the
index variables before applying make.pconsecutive, and especially check for NA values in the first
and last position for each individual in original data and, if so, maybe set those to some meaningful
begin/end value for the time series.
Value
An object of the same class as the input x, i.e. a pdata.frame, data.frame or a pseries which is made
time–consecutive based on the index variables. The returned data are sorted as a stacked time series.
Author(s)
Kevin Tappe
See Also
is.pconsecutive to check if data are consecutive; make.pbalanced to make data only balanced
(not consecutive).
punbalancedness for two measures of unbalancedness, pdim to check the dimensions of a ’pdata.frame’
(and other objects), pvar to check for individual and time variation of a ’pdata.frame’ (and other
objects), lag for lagged (and leading) values of a ’pseries’ object.
pseries, data.frame, pdata.frame.
Examples
# take data and make it non-consecutive
34 make.pconsecutive

nrow(Grunfeld) # 200 rows
is.pconsecutive(Grunfeld_missing_period) # check for consecutiveness
make.pconsecutive(Grunfeld_missing_period) # make it consecutiveness
# argument balanced:
# First, make data non-consecutive and unbalanced
# by deletion of 2nd time period (year 1936) for all individuals
# and more time periods for first individual only
Grunfeld_unbalanced <- Grunfeld[Grunfeld$year != 1936, ]
Grunfeld_unbalanced <- Grunfeld_unbalanced[-c(1,4), ]
all(is.pconsecutive(Grunfeld_unbalanced)) # FALSE
pdim(Grunfeld_unbalanced)$balanced # FALSE
g_consec_bal <- make.pconsecutive(Grunfeld_unbalanced, balanced = TRUE)

all(is.pconsecutive(g_consec_bal)) # TRUE
pdim(g_consec_bal)$balanced # TRUE
nrow(g_consec_bal) # 200 rows
head(g_consec_bal) # 1st individual: years 1935, 1936, 1939 are NA
g_consec <- make.pconsecutive(Grunfeld_unbalanced) # default: balanced = FALSE

all(is.pconsecutive(g_consec)) # TRUE
pdim(g_consec)$balanced # FALSE
nrow(g_consec) # 198 rows
head(g_consec) # 1st individual: years 1935, 1936 dropped, 1939 is NA
# NA in 1st, 3rd time period (years 1935, 1937) for first individual
Grunfeld_NA <- Grunfeld
Grunfeld_NA[c(1, 3), "year"] <- NA
g_NA <- make.pconsecutive(Grunfeld_NA)
head(g_NA) # 1936 is begin for 1st individual, 1937: NA for non-index vars
nrow(g_NA) # 199, year 1935 from original data is dropped
make.pconsecutive(Grunfeld_missing_period)
# pseries interface
make.pconsecutive(pGrunfeld_missing_period$inv)
# comparison to make.pbalanced (makes the data only balanced, not consecutive)

g_bal <- make.pbalanced(Grunfeld_unbalanced)
all(is.pconsecutive(g_bal)) # FALSE
pdim(g_bal)$balanced # TRUE
Males 35
Males Wages and Education of Young Males
Description
observation : individuals
Usage
data(Males)
Format
nr identifier
year year
school years of schooling
exper years of experience (computed as age-6-school)
union wage set by collective bargaining ?
ethn a factor with levels black, hisp, other
married married?
health health problem ?
wage log of hourly wage
industry a factor with 12 levels
occupation a factor with 9 levels
residence a factor with levels rural area, north east, northern central, south
Source
Journal of Applied Econometrics data archive http://qed.econ.queensu.ca/jae/1998-v13.2/
vella-verbeek/.
References
Vella, F. and M. Verbeek (1998) “Whose Wages Do Unions Raise? A Dynamic Model of Unionism
and Wage Rate Determination for Young Men”, Journal of Applied Econometrics, 13(2), pp. 163–
183.
Verbeek, M. (2004) A Guide to Modern Econometrics, John Wiley and Sons, chapter 10.
36 model.frame.pFormula
model.frame.pFormula model.frame and model.matrix for panel data
Description
Methods to create model frame and model matrix for panel data.
Usage

model.frame(formula, data, ...,
lhs = NULL, rhs = NULL)

model.matrix(object, data,
model = c("pooling","within","Between",
"between","mean","random","fd"),
effect = c("individual","time","twoways"),
rhs = 1,
theta = NULL, ...)

model.matrix(object, ...)
Arguments
object, formula
an object of class "pFormula" or an estimated model object of class "plm",
data a pdata.frame, see Details,
effect the effects introduced in the model, one of "individual", "time" or "twoways",
model one of "pooling", "within", "between", "random", "fd" and "ht",
theta the parameter for the transformation if model = "random",
lhs inherited from package Formula (see there),
rhs inherited from package Formula (see there),
Details
The lhs and rhs arguments are inherited from Formula, see there for more details.
The model.frame methods return a pdata.frame object suitable as an input to plm’s model.matrix.
The model.matrix methods builds a model matrix with transformations performed as specified by
the model and effect arguments (and theta if model = "random" is requested), in this case the
supplied data argument should be a model frame created by plm’s model.frame method. If not,
it is tried to construct the model frame from the data. Constructing the model frame first ensures
proper NA handling, see Examples.
mtest 37
Value
The model.frame methods return a pdata.frame.
The model.matrix methods return a matrix.
Author(s)
Yves Croissant
See Also
pmodel.response for (transformed) response variable.
pFormula, especially for coercing a formula to a pFormula.
Formula from package Formula, especially for the lhs and rhs arguments.
Examples
# First, make a pdata.frame
data(Grunfeld)
# then make a model frame from a pFormula and a pdata.frame

pform <- pFormula(inv ~ value + capital)
mf <- model.frame(pform, data = pGrunfeld)
# then construct the (transformed) model matrix (design matrix)

# from formula and model frame
modmat <- model.matrix(pform, data = mf, model = "within")
## retrieve model frame and model matrix from an estimated plm object
fe_model <- plm(pform, data = pGrunfeld, model = "within")
model.frame(fe_model)
model.matrix(fe_model)
# same as constructed before

all.equal(mf, model.frame(fe_model), check.attributes = FALSE) # TRUE
all.equal(modmat, model.matrix(fe_model), check.attributes = FALSE) # TRUE
mtest Arellano–Bond test of Serial Correlation
Description
Test of serial correlation for models estimated by GMM
Usage
mtest(object, order=1, vcov=NULL)
38 nobs
Arguments
object an object of class "pgmm",
order the order of the serial correlation (1 or 2),
vcov a matrix of covariance for the coefficients or a function to compute it.
Details
The Arellano–Bond test is a test of correlation based on the residuals of the estimation. By default,
the computation is done with the standard covariance matrix of the coefficients. A robust estimator
of this covariance matrix can be supplied with the vcov argument.
Value
Author(s)
Yves Croissant
References
Arellano, M. and Bond, S. (1991), Some Tests of Specification for Panel Data: Monte Carlo Evi-
dence and an Application to Employment Equations, The Review of Economic Studies, Vol. 58(2),
1991, pp. 227–297.
See Also
pgmm
Examples
ar <- pgmm(log(emp) ~ lag(log(emp), 1:2) + lag(log(wage), 0:1) +
lag(log(capital), 0:2) + lag(log(output), 0:2) | lag(log(emp), 2:99),
data = EmplUK, effect = "twoways", model = "twosteps")
mtest(ar, 1)
mtest(ar, 2, vcovHC)
nobs Extract Total Number of Observations Used in Estimated Panelmodel
Description
This function extracts the total number of ’observations’ from a fitted panel model.
nobs 39
Usage

nobs(object, ...)
Arguments
object a panelmodel object for which the number of total observations is to be ex-
tracted,
Details
The number of observations is usually the length of the residuals vector. Thus, nobs gives the
number of observations actually used by the estimation procedure. It is not necessarily the number
of observations of the model frame (number of rows in the model frame), because sometimes the
model frame is further reduced by the estimation procedure. This is e.g. the case for first–difference
models estimated by plm(..., model = "fd") where the model frame does not yet contain the
differences (see also Examples).
Value
A single number, normally an integer.
See Also
pdim
Examples
# estimate a panelmodel
z <- plm(log(gsp)~log(pcap)+log(pc)+log(emp)+unemp,data=Produc,
model="random", subset = gsp > 5000)
nobs(z) # total observations used in estimation

pdim(z)$nT$N # same information
pdim(z) # more information about the dimensions (no. of individuals and time periods)
# illustrate difference between nobs and pdim for first-difference model

fdmod <- plm(inv ~ value + capital, data = Grunfeld, model = "fd")
nobs(fdmod) # 190
pdim(fdmod)$nT$N # 200
40 Parity
Parity Purchasing Power Parity and other parity relationships
Description
A panel of 104 quarterly observations from 1973Q1 to 1998Q4
country : OECD
Usage
data(Parity)
Format
country country codes: a factor with 17 levels

time the quarter index, 1973Q1-1998Q4
ls log spot exchange rate vs. USD
lp log price level
is short term interest rate
il long term interest rate
ld log price differential vs. USA
uis U.S. short term interest rate
uil U.S. long term interest rate
Source
Coakley, J., Fuertes, A. M., and Smith, R. (2006) “Unobserved heterogeneity in panel time series
models”, Computational Statistics & Data Analysis, 50(9), 2361–2380.
References
Coakley, J., Fuertes, A. M., and Smith, R. (2006) “Unobserved heterogeneity in panel time series
models”, Computational Statistics & Data Analysis, 50(9), 2361–2380.
Driscoll, J. C., and Kraay, A. C. (1998). “Consistent covariance matrix estimation with spatially
dependent panel data”, Review of Economics and Statistics, 80(4), 549–560.
pbgtest 41
pbgtest Breusch–Godfrey Test for Panel Models
Description
Test of serial correlation for (the idiosyncratic component of) the errors in panel models.
Usage
pbgtest(x, ...)
pbgtest(x, order = NULL, type = c("Chisq", "F"), ...)
pbgtest(x, order = NULL, type = c("Chisq", "F"),
data, model=c("pooling", "random", "within"), ...)
Arguments
x an object of class "panelmodel" or of class "formula",
order an integer indicating the order of serial correlation to be tested for. NULL (default)
uses the minimum number of observations over the time dimension (see also
section Details below),
type type of test statistic to be calculated; either "Chisq" (default) for the Chi-
squared test statistic or "F" for the F test statistic,
data only relevant for formula interface: data set for which the respective panel model
(see model) is to be evaluated,
model only relevant for formula interface: compute test statistic for model pooling
(default), random, or within. When model is used, the data argument needs to
be passed as well,
... further arguments (see bgtest).
Details
This Lagrange multiplier test uses the auxiliary model on (quasi-)demeaned data taken from a model
of class plm which may be a pooling (default for formula interface), random or within model. It
performs a Breusch–Godfrey test (using bgtest from package lmtest) on the residuals of the
(quasi-)demeaned model, which should be serially uncorrelated under the null of no serial cor-
relation in idiosyncratic errors, as illustrated in Wooldridge (2002/2010). The function takes the
demeaned data, estimates the model and calls bgtest.
Unlike most other tests for serial correlation in panels, this one allows to choose the order of corre-
lation to test for.
Value
42 pbgtest
Note
The argument order defaults to the minimum number of observations over the time dimension,
while for lmtest::bgtest it defaults to 1.
Author(s)
Giovanni Millo
References
Breusch, T.S. (1978), “Testing for autocorrelation in dynamic linear models”, Australian Economic
Papers, 17(31), pp. 334–355.
Godfrey, L.G. (1978), “Testing against general autoregressive and moving average error models
when the regressors include lagged dependent variables”, Econometrica, 46(6), pp. 1293–1301.
Wooldridge, J.M. (2002) Econometric Analysis of Cross-Section and Panel Data, MIT Press, pp.
288–291.
Wooldridge, J.M. (2010) Econometric analysis of cross-section and Panel Data, 2nd ed., MIT Press,
pp. 328–334.
Wooldridge, J.M. (2013) Introductory Econometrics: A Modern Approach, 5th ed., South-Western
(Cengage Learning), Sec. 12.2, pp. 421–422.
See Also
pdwtest for the analogous panel Durbin–Watson test, bgtest for the Breusch–Godfrey test for
serial correlation in the linear model. pbltest, pbsytest, pwartest and pwfdtest for other serial
correlation tests for panel models. For the original test in package lmtest see bgtest.
Examples
g <- plm(inv ~ value + capital, data = Grunfeld, model = "random")
# panelmodel interface
pbgtest(g)
pbgtest(g, order = 4)
# formula interface
pbgtest(inv ~ value + capital, data = Grunfeld, model = "random")
# F test statistic (instead of default type="Chisq")

pbgtest(g, type="F")
pbgtest(inv ~ value + capital, data = Grunfeld, model = "random", type="F")
# same output as lmtest::bgtest [see Note]

pbgtest(g, order = 1)
lmtest::bgtest(g, order = 1)
pbltest 43
pbltest Baltagi and Li Serial Dependence Test For Random Effects Models
Description
Baltagi and Li (1995)’s Lagrange multiplier test for AR(1) or MA(1) idiosyncratic errors in panel
models with random effects.
Usage
pbltest(x, data,
alternative = c("twosided", "onesided"), index = NULL, ...)
pbltest(x, alternative = c("twosided", "onesided"), ...)
Arguments
x a model formula or an estimated random–effects model of class plm ,
data for the formula interface only: a data.frame,
alternative one of "twosided", "onesided". Selects either HA : ρ 6= 0 or HA : ρ = 0
(i.e., the Normal or the Chi-squared version of the test),
index the index of the data.frame,
Details
This is a Lagrange multiplier test for the null of no serial correlation, against the alternative of either
an AR(1) or an MA(1) process, in the idiosyncratic component of the error term in a random effects
panel model (as the analytical expression of the test turns out to be the same under both alternatives,
see Baltagi and Li (1995, 1997)). The alternative argument, defaulting to twosided, allows
testing for positive serial correlation only, if set to onesided.
Value
Author(s)
Giovanni Millo
References
Baltagi, B.H. and Li, Q. (1995) Testing AR(1) against MA(1) disturbances in an error component
model, Journal of Econometrics 68, pp. 133–151.
Baltagi, B.H. and Li, Q. (1997) Monte Carlo results on pure and pretest estimators of an error
component model with autocorrelated disturbances, Annales d’economie et de statistique 48, pp.
69–82.
44 pbsytest
See Also
pdwtest, bgtest, pbsytest, pwartest and pwfdtest for other serial correlation tests for panel
models.
Examples
# formula interface
pbltest(inv ~ value + capital, data = Grunfeld)
# plm interface
re_mod <- plm(inv ~ value + capital, data = Grunfeld, model = "random")
pbltest(re_mod)
pbltest(re_mod, alternative = "onesided")
pbsytest Bera, Sosa-Escudero and Yoon Locally–Robust Lagrange Multiplier

Tests for Panel Models
Description
Test for residual serial correlation (or individual random effects) locally robust vs. individual ran-
dom effects (serial correlation) for panel models and joint test by Baltagi and Li.
Usage
pbsytest(x, ...)
pbsytest(x, test = c("ar","re","j"), ...)
pbsytest(x, data, ..., test = c("ar","re","j"))
Arguments
x an object of class "formula" or of class panelmodel,

data a data.frame,
test a character string indicating which test to perform: first-order serial correlation
(ar), random effects (re) or joint test for either of them (j).
pcce 45
Details
These Lagrange multiplier tests are robust vs. local misspecification of the alternative hypothesis,
i.e. they test the null of serially uncorrelated residuals against AR(1) residuals in a pooling model,
allowing for local departures from the assumption of no random effects; or they test the null of no
random effects allowing for local departures from the assumption of no serial correlation in residu-
als. They use only the residuals of the pooled OLS model and correct for local misspecification as
outlined in Bera et al. (2001).
The joint test is due to Baltagi and Li (1991) and is added for convenience under this same interface.
Value
Author(s)
Giovanni Millo
References
Bera, A.K., Sosa-Escudero, W. and Yoon, M. (2001), Tests for the error component model in the
presence of local misspecification, Journal of Econometrics, 101(1), pp. 1–23.
Baltagi, B. and Li, Q. (1991), A joint test for serial correlation and random individual effects,
Statistics and Probability Letters, 11(3), pp. 277–280.
See Also
plmtest for individual and/or time random effects tests based on a correctly specified model;
pbltest, pbgtest and pdwtest for serial correlation tests in random effects models.
Examples
## Example in Bera et al.
## Bera et al. use a subset of the original Grunfeld data,
## so results are slightly different here
## default is AR testing
pbsytest(inv ~ value + capital, data = Grunfeld, index = c("firm","year"))
pbsytest(inv ~ value + capital, data = Grunfeld, test="re", index = c("firm","year"))
pbsytest(inv ~ value + capital, data = Grunfeld, test="j", index = c("firm","year"))
pcce Common Correlated Effects estimators
Description
Common Correlated Effects Mean Groups (CCEMG) and Pooled (CCEP) estimators for panel data
with common factors (balanced or unbalanced)
46 pcce
Usage
pcce(formula, data, subset, na.action,
model=c("mg", "p"),
index = NULL, trend = FALSE, ...)
## S3 method for class 'pcce'
## S3 method for class 'summary.pcce'
print(x,digits = max(3, getOption("digits") -
Arguments
formula a symbolic description of the model to be estimated,
object, x an object of class pcce,
data a data.frame,
subset see lm,
na.action see lm,
model one of c("mg","p"), selects Mean Groups vs. Pooled CCE model,
index the indexes, see pdata.frame,
trend logical specifying whether an individual-specific trend has to be included,
digits digits,
Details
pcce is a function for the estimation of linear panel models by the Common Correlated Effects Mean
Groups or Pooled estimator, consistent under the hypothesis of unobserved common factors and
idiosyncratic factor loadings; CCE estimators works by augmenting the model by cross-sectional
averages of the dependent variable and regressors in order to account for the common factors, and
adding individual intercepts and possibly trends.
Value
An object of class c("pcce","panelmodel") containing:
coefficients the vector of coefficients,
residuals the vector of (defactored) residuals,
stdres the vector of (raw) residuals,
tr.model the transformed data after projection on H,
fitted.values the vector of fitted.values,
vcov the covariance matrix of the coefficients,
df.residual degrees of freedom of the residuals,
model a data.frame containing the variables used for the estimation,
pcdtest 47
call the call,

sigma always NULL, sigma is here only for compatibility reasons (to allow using the
same summary and print methods as pggls),
indcoef the matrix of individual coefficients from separate time series regressions.
Author(s)
Giovanni Millo
References
G. Kapetanios, M. Hashem Pesaran, T. Yamagata (2011), Panels with non-stationary multifactor

error structures, Journal of Econometrics, 160(2), pp. 326–348.
Examples
ccepmod <- pcce(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp, data = Produc, model="p")
summary(ccepmod)
pcdtest Tests of cross-section dependence for panel models
Description
Pesaran’s CD or Breusch–Pagan’s LM (local or global) tests for cross sectional dependence in panel
models
Usage
pcdtest(x, ...)
pcdtest(x, data, index = NULL, model = NULL,
test = c("cd", "sclm", "lm", "rho", "absrho"),
w = NULL, ...)
pcdtest(x, test = c("cd", "sclm", "lm", "rho", "absrho"),
w = NULL, ...)
pcdtest(x, test = c("cd", "sclm", "lm", "rho", "absrho"),
w = NULL, ...)
48 pcdtest
Arguments
x an object of class formula, panelmodel, or pseries (depending on the respec-

tive interface) describing the model to be tested
data a data.frame
index an optional numerical index, in case data has to be formatted by plm.data
model an optional character string indicating which type of model to estimate; if left to
NULL, the original heterogeneous specification of Pesaran is used
test the type of test statistic to be returned. One of
• "cd" for Pesaran’s CD statistic,
• "lm" for Breusch and Pagan’s original LM statistic,
• "sclm" for the scaled version of Breusch and Pagan’s LM statistic, or
• "rho" for the average correlation coefficient,
• "absrho" for the average absolute correlation coefficient,
w either NULL (default) for the global tests or - for the local versions of the statis-
tics - a n x n matrix describing proximity between individuals, with wi j = a
where a is any number such that as.logical(a)==TRUE, if i, j are neighbours,
0 or any number b such that as.logical(b)==FALSE elsewhere. Only the lower
triangular part (without diagonal) of w after coercing by as.logical() is eval-
uated for neighbouring information (but w can be symmetric). See also Details
and Examples.
... further arguments to be passed on to plm, such as e.g. effect or random.method
Details
These tests are originally meant to use the residuals of separate estimation of one time–series regres-
sion for each cross-sectional unit in order to check for cross–sectional dependence. If a different
model specification (within, random, . . . ) is assumed consistent, one can resort to its residuals for
testing (which is common, e.g., when the time dimension’s length is insufficient for estimating the
heterogeneous model). If the time dimension is insufficient and model=NULL, the function defaults
to estimation of a within model and issues a warning. The main argument of this function may be
either a model of class panelmodel or a formula and dataframe; in the second case, unless model
is set to NULL, all usual parameters relative to the estimation of a plm model may be passed on.
The test is compatible with any consistent panelmodel for the data at hand, with any specification
of effect. E.g., specifying effect=''time'' or effect=''twoways'' allows to test for resid-
ual cross-sectional dependence after the introduction of time fixed effects to account for common
shocks.
A local version of either test can be computed by supplying a proximity matrix (elements coercible
to logical) with argument w which provides information on whether any pair of individuals are
neighbours or not. If w is supplied, only neighbouring pairs will be used in computing the test; else,
w will default to NULL and all observations will be used. The matrix need not be binary, so commonly
used “row–standardized” matrices can be employed as well. nb objects from spdep must instead be
transformed into matrices by spdep’s function nb2mat before using.
The methods implemented are suitable also for unbalanced panels.
pcdtest 49
Pesaran’s CD test (test="cd"), Breusch and Pagan’s LM test (test="lm"), and it’s scaled ver-
sion (test="sclm") are all described in Pesaran (2004) (and complemented by Pesaran (2015)).
Breusch/Pagan (1980) is the original source for the LM test.
The test on a pseries is the same as a test on a pooled regression model of that variable on a con-
stant, i.e. pcdtest(some_pseries) is equivalent to pcdtest(plm(some_var ~ 1, data = some_pdata.frame, model = "
and also equivalent to pcdtest(some_var ~ 1, data = some_data), where some_var is the vari-
able name in the data which corresponds to some_pseries.
Value
References
Breusch, T.S. and A.R. Pagan (1980), The Lagrange multiplier test and its applications to model
specification in econometrics, Review of Economic Studies, 47(1), pp. 239–253.
Pesaran, M.H. (2004), General Diagnostic Tests for Cross Section Dependence in Panels, CESifo
Working Paper 1229.
Pesaran, M.H. (2015), Testing Weak Cross–Sectional Dependence in Large Panels, Econometric
Reviews, 34(6-10), pp. 1089–1117.
Examples
## test on heterogeneous model (separate time series regressions)
pcdtest(inv ~ value + capital, data = Grunfeld,
index = c("firm", "year"))
## test on two-way fixed effects homogeneous model

pcdtest(inv ~ value + capital, data = Grunfeld, model = "within",
effect = "twoways", index = c("firm", "year"))
## test on panelmodel object

g <- plm(inv ~ value + capital, data = Grunfeld, index = c("firm", "year"))
pcdtest(g)
## scaled LM test
pcdtest(g, test = "sclm")
## test on pseries
pcdtest(pGrunfeld$value)
## local test
## define neighbours for individual 2: 1, 3, 4, 5 in lower triangular matrix
w <- matrix(0, ncol= 10, nrow=10)
w[2,1] <- w[3,2] <- w[4,2] <- w[5,2] <- 1
pcdtest(g, w = w)
50 pdata.frame
pdata.frame data.frame for panel data
Description
An object of class ’pdata.frame’ is a data.frame with an index attribute that describes its individual
and time dimensions.
Usage
pdata.frame(x, index = NULL, drop.index = FALSE, row.names = TRUE,
stringsAsFactors = default.stringsAsFactors())
x[i, j, drop]
x[[y]]
x$y
print(x, ...)
as.data.frame(x, row.names = NULL, optional = FALSE, ...)
as.list(x, keep.attributes = FALSE, ...)
Arguments
x a data.frame for the pdata.frame function and a pdata.frame for the meth-
ods,
i see Extract,
j see Extract,
y one of the columns of the data.frame,
index this argument indicates the individual and time indexes. See Details,
drop see Extract,
drop.index logical, indicates whether the indexes are to be excluded from the resulting
pdata.frame,
optional see as.data.frame,
row.names NULL or logical, indicates whether “fancy” row names (a combination of indi-
vidual index and time index) are to be added to the returned (p)data.frame (NULL
and FALSE have the same meaning),
stringsAsFactors
logical, indicating whether character vectors are to be converted to factors,
keep.attributes
logical, only for as.list, indicating the elements of the returned list should have
the pdata.frame’s attributes added (default: FALSE),
... further arguments
pdata.frame 51
Details
The index argument indicates the dimensions of the panel. It can be:
• a vector of two character strings which contains the names of the individual and of the time
indexes,
• a character string which is the name of the individual index variable. In this case, the time
index is created automatically and a new variable called “time” is added, assuming consecutive
and ascending time periods in the order of the original data,
• an integer, the number of individuals. In this case, the data need to be a balanced panel and be
organized as a stacked time series (successive blocks of individuals, each block being a time
series for the respective individual) assuming consecutive and ascending time periods in the
order of the original data. Two new variables are added: “time” and “id” which contain the
individual and the time indexes.
The "[[" and "$" extract a series from the pdata.frame. The "index" attribute is then added to
the series and a class attribute "pseries" is added. The "[" method behaves as for data.frame,
except that the extraction is also applied to the index attribute. as.data.frame removes the index
from the pdata.frame and adds it to each column.
as.list behaves by default identical to as.list.data.frame which means it drops the attributes
specific to a pdata.frame; if a list of pseries is wanted, the attribute keep.attributes can to be set
to TRUE. This also makes lapply work as expected on a pdata.frame (see also Examples).
Value
a pdata.frame object: this is a data.frame with an index attribute which is a data.frame with
two variables, the individual and the time indexes, both being factors. The resulting pdata.frame is
sorted by the individual index, then by the time index. Any constant columns and all-NA columns
are dropped.
Author(s)
Yves Croissant
See Also
pdim to check the dimensions of a ’pdata.frame’ (and other objects), pvar to check for each vari-
able if it varies cross-sectionally and over time. To check if the time periods are consecutive per
individual, see is.pconsecutive.
Examples
# Gasoline contains two variables which are individual and time indexes
data("Gasoline", package = "plm")
Gas <- pdata.frame(Gasoline, index = c("country", "year"), drop = TRUE)
# Hedonic is an unbalanced panel, townid is the individual index

data("Hedonic", package = "plm")
Hed <- pdata.frame(Hedonic, index = "townid", row.names = FALSE)
52 pdim
# In case of balanced panel, it is sufficient to give number of individuals

# data set 'Wages' is organized as a stacked time series
data("Wages", package = "plm")
Wag <- pdata.frame(Wages, 595)
# lapply on a pdata.frame by making it a list of pseries first

lapply(as.list(Wag[ , c("ed", "lwage")], keep.attributes = TRUE), lag)
pdim Check for the Dimensions of the Panel
Description
This function checks the number of individuals and time observations in the panel and whether it is
balanced or not.
Usage
pdim(x, ...)
pdim(x, index = NULL, ...)
pdim(x, ...)
pdim(x, ...)
pdim(x, ...)
pdim(x, ...)
Arguments
x a data.frame, a pdata.frame, a pseries, a panelmodel, or a pgmm object,
index see pdata.frame,
Details
pdim is called by the estimation functions and can be also used stand-alone.
Value
An object of class pdim containing the following elements:
nT a list containing n, the number of individuals, T, the number of time observa-

tions, N the total number of observations,
pdim 53
Tint a list containing two vectors (of type integer): Ti gives the number of observa-
tions for each individual and nt gives the number of individuals observed for
each period,
balanced a logical value: TRUE for a balanced panel, FALSE for an unbalanced panel,
panel.names a list of character vectors: id.names contains the names of each individual and
time.names contains the names of each period.
Note
Calling pdim on an estimated panelmodel object and on the corresponding (p)data.frame used for
this estimation does not necessarily yield the same result. When called on an estimated panelmodel,
the number of observations (individual, time) actually used for model estimation are taken into ac-
count. When called on a (p)data.frame, the rows in the (p)data.frame are considered, disre-
garding any NA values in the dependent or independent variable(s) which would be dropped during
model estimation.
Author(s)
Yves Croissant
See Also
is.pbalanced to just determine balancedness of data (slightly faster than pdim),
punbalancedness for measures of unbalancedness,
nobs, pdata.frame,
pvar to check for each variable if it varies cross-sectionally and over time.
Examples
# There are 595 individuals
pdim(Wages, 595)
# and are the first two variables
data("Gasoline", package="plm")
pdim(Gasoline)

pdim(Hedonic, "townid")
# An example of the panelmodel method

z <- plm(log(gsp)~log(pcap)+log(pc)+log(emp)+unemp,data=Produc,
model="random", subset = gsp > 5000)
pdim(z)
54 pdwtest
pdwtest Durbin–Watson Test for Panel Models
Description
Test of serial correlation for (the idiosyncratic component of) the errors in panel models.
Usage
pdwtest(x, ...)
pdwtest(x, ...)
pdwtest(x, data, ...)
Arguments
x an object of class "panelmodel" or of class "formula",
data a data.frame,
... further arguments to be passed on to dwtest.
Details
This Durbin–Watson test uses the auxiliary model on (quasi-)demeaned data taken from a model
of class plm which may be a pooling (the default), random or within model. It performs a dw
test (using dwtest from package lmtest) on the residuals of the (quasi-)demeaned model, which
should be serially uncorrelated under the null of no serial correlation in idiosyncratic errors. The
function takes the demeaned data, estimates the model and calls dwtest.
Value
Author(s)
Giovanni Millo
References
Baltagi, B.H. (2005) Econometric Analysis of Panel Data, 3rd. ed., Wiley, p. 98.
Wooldridge, J.M. (2002) Econometric Analysis of Cross-Section and Panel Data, MIT Press, p.
288.
See Also
pbgtest for the analogous Breusch–Godfrey test, dwtest for the Breusch–Godfrey test for serial
correlation in the linear model. pbltest, pbsytest, pwartest and pwfdtest for other serial cor-
relation tests for panel models.
pFormula 55
Examples
g <- plm(inv ~ value + capital, data = Grunfeld, model="random")
pdwtest(g)
pdwtest(g, alternative="two.sided")
## formula interface
pdwtest(inv ~ value + capital, data=Grunfeld, model="random")
pFormula pFormula: An extended Formula interface for panel data
Description
pFormula is a Formula object used in the plm package.
Usage
pFormula(object)
as.Formula(x, ...)
Arguments
object an object of class "formula", the formula to be coerced to class "pFormula",
x an object of class "pFormula", to be coerced to class{"Formula"},
Value
For pFormula, the return value is an object of class c("pFormula", "Formula", "formula").
For as.Formula, the return value is an object of class c("Formula", "formula").
Author(s)
Yves Croissant
See Also
plm’s model.frame and model.matrix to create a model frame for panel data and a model matrix
with data transformations applied, respectively.
plm’s pmodel.response for (transformed) response variable.
Formula from package Formula.
56 pFtest
Examples
data(Grunfeld)

# then construct the (transformed) model matrix (design matrix)

# from formula and model frame
modmat <- model.matrix(pform, data = mf, model = "within")
pFtest F Test for Individual and/or Time Effects
Description
Test of individual and/or time effects based on the comparison of the within and the pooling
model.
Usage
pFtest(x, ...)
pFtest(x, z, ...)
pFtest(x, data, ...)
Arguments
x an object of class "plm" or of class "formula",
z an object of class "plm",
data a data.frame,
Details
For the plm method, the argument of this function is two plm objects, the first being a within model,
the second a pooling model. The effects tested are either individual, time or twoways, depending
on the effects introduced in the within model.
Value
pggls 57
Author(s)
Yves Croissant
See Also
plmtest for Lagrange multipliers tests of individuals and/or time effects.
Examples
data("Grunfeld", package="plm")
gp <- plm(inv ~ value + capital, data = Grunfeld, model = "pooling")
gi <- plm(inv ~ value + capital, data = Grunfeld,
effect = "individual", model = "within")
gt <- plm(inv ~ value + capital, data = Grunfeld,
effect = "time", model = "within")
gd <- plm(inv ~ value + capital, data = Grunfeld,
effect = "twoways", model = "within")
pFtest(gi, gp)
pFtest(gt, gp)
pFtest(gd, gp)
pFtest(inv ~ value + capital, data = Grunfeld, effect = "twoways")
pggls General FGLS Estimators
Description
General FGLS estimators for panel data (balanced or unbalanced)
Usage
pggls(formula, data, subset, na.action, effect = c("individual", "time"),
model = c("within","random","pooling","fd"),
index = NULL, ...)
## S3 method for class 'pggls'
## S3 method for class 'summary.pggls'
print(x,digits = max(3, getOption("digits") - 2),
Arguments
object, x an object of class pggls,
data a data.frame,
subset see lm,
na.action see lm,
58 pggls
effect the effects introduced in the model, one of "individual" or "time",

model one of "within", "pooling", "random" or "fd",
digits digits,
Details
pggls is a function for the estimation of linear panel models by general feasible generalized least
squares, either with or without fixed effects. General FGLS is based on a two-step estimation pro-
cess: first a model is estimated by OLS (pooling), fixed effects (within) or first differences (fd),
then its residuals are used to estimate an error covariance matrix for use in a feasible-GLS analysis.
This framework allows the error covariance structure inside every group (if effect="individual",
else symmetric) of observations to be fully unrestricted and is therefore robust against any type of
intragroup heteroskedasticity and serial correlation. Conversely, this structure is assumed identi-
cal across groups and thus general FGLS estimation is inefficient under groupwise heteroskedas-
ticity. Note also that this method requires estimation of T (T + 1)/2 variance parameters, thus
efficiency requires N > > T (if effect="individual", else the opposite). The model="random"
and model="pooling" arguments both produce an unrestricted FGLS model as in Wooldridge,
Ch. 10, although the former is deprecated and included only for retro–compatibility reasons.
If model="within" (the default) then a FEGLS (fixed effects GLS, see ibid.) is estimated; if
model="fd" a FDGLS (first-difference GLS).
Value
An object of class c("pggls","panelmodel") containing:

residuals the vector of residuals,
call the call,
sigma the estimated intragroup (or cross-sectional, if effect="time") covariance of
errors,
Author(s)
Giovanni Millo
pgmm 59
References
Kiefer, N. M. (1980) Estimation of Fixed Effects Models for Time Series of Cross-Sections with
Arbitrary Intertemporal Covariance, Journal of Econometrics, 14(2), pp. 195–202.
Im, K. S. and Ahn, S. C. and Schmidt, P. and Wooldridge, J. M. (1999) Efficient Estimation of Panel
Data Models with Strictly Exogenous Explanatory Variables, Journal of Econometrics, 93(1), pp.
177–201.
Wooldridge, J. M. (2002) Econometric Analysis of Cross Section and Panel Data, MIT Press.
Wooldridge, J. M. (2010) Econometric analysis of cross-section and Panel Data, 2nd ed., MIT
Press.
Examples
zz_wi <- pggls(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp, data = Produc, model = "within")
summary(zz_wi)
zz_pool <- pggls(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp,

data = Produc, model = "pooling")
summary(zz_pool)
zz_fd <- pggls(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp,

data = Produc, model = "fd")
summary(zz_fd)
pgmm Generalized Method of Moments (GMM) Estimation for Panel Data
Description
Generalized method of moments estimation for static or dynamic models with panel data.
Usage
pgmm(formula, data, subset, na.action,
effect = c("twoways", "individual"),
model = c("onestep", "twosteps"),
collapse = FALSE,
lost.ts = NULL,
transformation = c("d", "ld"),
fsm = NULL, index = NULL, ...)
summary(object, robust, time.dummies = FALSE, ...)
## S3 method for class 'summary.pgmm'
60 pgmm
Arguments
formula a symbolic description for the model to be estimated. The preferred interface is
now to indicate a multi–part formula, the first two parts describing the covariates
and the gmm instruments and, if any, the third part the ’normal’ instruments,
object,x an object of class "pgmm",
data a data.frame (neither factors nor character vectors will be accepted in data.frame),
subset see lm,
na.action see lm,
effect the effects introduced in the model, one of "twoways" (the default) or "individual",
model one of "onestep" (the default) or "twosteps",
collapse if TRUE, the gmm instruments are collapsed,
lost.ts the number of lost time series: if NULL, this is automatically computed. Other-
wise, it can be defined by the user as a numeric vector of length 1 or 2. The first
element is the number of lost time series in the model in difference, the second
one in the model in level. If the second element is missing, it is set to the first
one minus one,
transformation the kind of transformation to apply to the model: either "d" (the default value)
for the “difference GMM” model or "ld" for the “system GMM”,
fsm the matrix for the one step estimator: one of "I" (identity matrix) or "G" (=
D0 D where D is the first–difference operator) if transformation="d", one of
"GI" or "full" if transformation="ld",
index the indexes,
digits digits,
robust if TRUE, robust inference is performed in the summary,
time.dummies if TRUE, the estimated coefficients of time dummies are present in the table of
coefficients,
Details
pgmm estimates a model for panel data with a generalized method of moments (GMM) estimator.
The description of the model to estimate is provided with a multi–part formula which is (or which
is coerced to) a Formula object. The first right–hand side part describes the covariates. The second
one, which is mandatory, describes the gmm instruments. The third one, which is optional, describes
the ’normal’ instruments. By default, all the variables of the model which are not used as GMM
instruments are used as normal instruments with the same lag structure as the one specified in the
model.
y~lag(y, 1:2)+lag(x1, 0:1)+lag(x2, 0:2) | lag(y, 2:99) is similar to y~lag(y, 1:2)+lag(x1, 0:1)+lag(x2, 0:
lag(x1, 0:1)+lag(x2, 0:2) and indicates that all lags from 2 of y is used as gmm instruments.
transformation indicates how the model should be transformed for the estimation. "d" gives the
“difference GMM” model (see Arellano and Bond (1991)), "ld" the “system GMM” model (see
Blundell and Bond (1998)).
pgmm 61
pgmm is an attempt to adapt GMM estimators available within the DPD library for GAUSS (see Arel-
lano and Bond (1998)) and Ox (see Doornik, Arellano and Bond (2006)) and within the xtabond2
library for Stata (see Roodman (2009)).
Value
An object of class c("pgmm","panelmodel"), which has the following elements:
coefficients the vector (or the list for fixed effects) of coefficients,
fitted.values the vector of fitted values,
model a list containing the variables used for the estimation for each individual,
W a list containing the instruments for each individual (two lists in case of “sys–
GMM”),
A1 the weighting matrix for the one–step estimator,
A2 the weighting matrix for the two–steps estimator,
call the call.
It has print, summary and print.summary methods.
Author(s)
Yves Croissant
References
Arellano, M. and Bond, S. (1991) Some Tests of Specification for Panel Data: Monte Carlo Evi-
dence and an Application to Employment Equations, The Review of Economic Studies, vol. 58(2),
1991, pp. 227–297.
Arellano, M. and Bond, S. (1998) Dynamic Panel Data Estimation Using DPD98 for GAUSS: A
Guide for Users.
Blundell, R. and Bond, S. (1998) Initial Conditions and Moment Restrictions in Dynamic Panel
Data Models, Journal of Econometrics, vol. 87(1), pp. 115–143.
Doornik, J., Arellano, M. and Bond, S. (2006) Panel Data Estimation using DPD for Ox. http:
//www.doornik.com/download/oxmetrics7/Ox_Packages/dpd.pdf
Roodman, D. (2009) How to do xtabond2: An Introduction to difference and system GMM in
Stata, Stata Journal, vol. 9(1), pp. 86–136. http://www.stata-journal.com/article.html?
article=st0159.
See Also
dynformula for dynamic formulas, sargan for Sargan tests and mtest for Arellano–Bond’s tests
of serial correlation.
62 pht
Examples
## Arellano and Bond (1991), table 4 col. b

z1 <- pgmm(log(emp) ~ lag(log(emp), 1:2) + lag(log(wage), 0:1)
+ log(capital) + lag(log(output), 0:1) | lag(log(emp), 2:99),
summary(z1)
## Blundell and Bond (1998) table 4 (cf. DPD for OX p. 12 col. 4)

z2 <- pgmm(log(emp) ~ lag(log(emp), 1)+ lag(log(wage), 0:1) +
lag(log(capital), 0:1) | lag(log(emp), 2:99) +
lag(log(wage), 2:99) + lag(log(capital), 2:99),
data = EmplUK, effect = "twoways", model = "onestep",
transformation = "ld")
summary(z2, robust = TRUE)
## Not run:
## Same with the old formula or dynformula interface
## Arellano and Bond (1991), table 4, col. b
z1 <- pgmm(log(emp) ~ log(wage) + log(capital) + log(output),
lag.form = list(2,1,0,1), data = EmplUK,
effect = "twoways", model = "twosteps",
gmm.inst = ~log(emp), lag.gmm = list(c(2,99)))
summary(z1)
## Blundell and Bond (1998) table 4 (cf DPD for OX p. 12 col. 4)

z2 <- pgmm(dynformula(log(emp) ~ log(wage) + log(capital), list(1,1,1)),
data = EmplUK, effect = "twoways", model = "onestep",
gmm.inst = ~log(emp) + log(wage) + log(capital),
lag.gmm = c(2,99), transformation = "ld")
summary(z2, robust = TRUE)
## End(Not run)
pht Hausman–Taylor Estimator for Panel Data
Description
The Hausman–Taylor estimator is an instrumental variable estimator without external instruments.
Usage
pht(formula, data, subset, na.action, model = c("ht", "am", "bmc"), index = NULL, ...)
## S3 method for class 'pht'
## S3 method for class 'summary.pht'
width = getOption("width"), subset = NULL, ...)
pht 63
Arguments

object,x an object of class "plm",
data a data.frame,
subset see lm for "plm", a character or numeric vector indicating a subset of the table
of coefficient to be printed for "print.summary.plm",
na.action see lm,
model one of "ht" for Hausman–Taylor, "am" for Amemiya–MaCurdy and "bms" for
Breusch–Mizon–Schmidt,
index the indexes,
digits digits,
Details
pht estimates panels models using the Hausman–Taylor estimator, Amemiya–MaCurdy estimator,
or Breusch–Mizon–Schmidt estimator, depending on the argument model. The model is specified
as a two–part formula, the second part containing the exogenous variables.
Value
An object of class c("pht", "plm", "panelmodel").

A "pht" object contains the same elements as plm, with a further argument called varlist which
describes the typology of the variables. It has summary and print.summary methods.
Author(s)
Yves Croissant
References
Amemiya, T. and MaCurdy, T.E. (1986) Instrumental–variable estimation of an error components

model, Econometrica, 54(4), pp. 869–880.
Breusch, T.S., Mizon, G.E. and Schmidt, P. (1989) Efficient estimation using panel data, Economet-
rica, 57(3), pp. 695–700.
Hausman, J.A. and Taylor W.E. (1981) Panel data and unobservable individual effects, Economet-
rica, 49(6), pp. 1377–1398.
64 phtest
Examples
# replicates Baltagi (2005, 2013), table 7.4

ht <- pht(lwage ~ wks + south + smsa + married + exp + I(exp^2) +
bluecol + ind + union + sex + black + ed |
sex + black + bluecol + south + smsa + ind,
data = Wages, model = "ht", index = 595)
summary(ht)
phtest Hausman Test for Panel Models
Description
Specification test for panel models.
Usage
phtest(x, ...)
phtest(x, x2, ...)
phtest(x, data, model = c("within", "random"),
method = c("chisq", "aux"),
index = NULL, vcov = NULL, ...)
Arguments
x an object of class "panelmodel" or "formula",

x2 an object of class "panelmodel",
model a character vector containing the names of two models (length(model) must be
2),
data a data.frame,
method one of "chisq" or "aux",
index an optional vector of index variables,
vcov an optional covariance function,
... further arguments to be passed on. For the formula method, place argument
effect here to compare e.g. twoway models (effect = "twoways") Note:
Argument effect is not respected in the panelmodel method.
phtest 65
Details
The Hausman test (sometimes also called Durbin–Wu–Hausman test) is based on the difference of
the vectors of coefficients of two different models. The panelmodel method computes the orig-
inal version of the test based on a quadratic form (Hausman (1978)). The formula method, if
method="chisq" (default), computes the original version of the test based on a quadratic form; if
method="aux" then the auxiliary-regression-based version in Wooldridge (2010, Sec. 10.7.3.) is
computed instead. Only the latter can be robustified by specifying a robust covariance estimator as
a function through the argument vcov (see Examples).
The equivalent tests in the one-way case using a between model (either "within vs. between" or
"random vs. between") (see Hausman/Taylor (1981) or Baltagi (2013), Sec. 4.3) can also be per-
formed by phtest, but only for test = "chisq", not for the regression-based test. NB: These
equivalent tests using the between model do not extend to the two-ways case. (There are, how-
ever, some other equivalent tests, see Kang (1985) or Baltagi (2013), Sec. 4.3.7), but those are
unsupported by phtest.)
Value
Author(s)
Yves Croissant, Giovanni Millo
References
Hausman, J.A. (1978), Specification tests in econometrics, Econometrica, 46(6), pp. 1251–1271.
Hausman, J.A./Taylor, W.E. (1981), Panel data and unobservable individual effects, Econometrica,
49(6), pp. 1377–1398.
Kang, Suk (1985), A note on the equivalence of specification tests in the two-factor multivariate
variance components model, Journal of Econometrics, 28(2), pp. 193–203.
Wooldridge, Jeffrey M. (2010), Econometric analysis of cross-section and panel data, 2nd ed., MIT
Press, Sec. 10.7.3., pp. 328–334.
Baltagi, Badi H. (2013), Econometric Analysis of Panel Data, 5th ed., John Wiley and Sons., Sec.
4.3.
Examples
form <- lgaspcar ~ lincomep + lrpmg + lcarpcap
wi <- plm(form, data = Gasoline, model = "within")
re <- plm(form, data = Gasoline, model = "random")
phtest(wi, re)
phtest(form, data = Gasoline)
phtest(form, data = Gasoline, method = "aux")
# robust Hausman test (regression-based)

phtest(form, data = Gasoline, method = "aux", vcov = vcovHC)
66 piest
# robust Hausman test with vcov supplied as a

# function and additional parameters
phtest(form, data = Gasoline, method = "aux",
vcov = function(x) vcovHC(x, method="white2", type="HC3"))
piest Chamberlain estimator and test for fixed effects
Description
General estimator useful for testing the within specification
Usage
piest(formula, data, subset, na.action, index = NULL, robust = TRUE, ...)
## S3 method for class 'piest'
## S3 method for class 'summary.piest'
print(x, ...)
Arguments
object,x an object of class "plm",
data a data.frame,
subset see lm,
na.action see lm,
index the indexes,
robust if FALSE, the error as assumed to be spherical, otherwise, a robust estimation of
the covariance matrix is computed,
Details
The Chamberlain method consists on using the covariates of all the periods as regressors. It allows
to test the within specification.
Value
An object of class "piest".
Author(s)
Yves Croissant
plm 67
References
Chamberlain, G. (1982) Multivariate regression for panel data, Journal of Econometrics, 18(1), pp.
5–46.
Examples
data("RiceFarms", package = "plm")
pirice <- piest(log(goutput) ~ log(seed) + log(totlabor) + log(size), RiceFarms, index = "id")
summary(pirice)
plm Panel Data Estimators
Description
Linear models for panel data estimated using the lm function on transformed data.
Usage
plm(formula, data, subset, na.action, effect = c("individual", "time", "twoways"),
model = c("within", "random", "ht", "between", "pooling", "fd"),
random.method = c("swar", "walhus", "amemiya", "nerlove", "kinla"),
random.dfcor = NULL,
inst.method = c("bvk", "baltagi", "am", "bmc"), restrict.matrix = NULL,
restrict.rhs = NULL, index = NULL, ...)
print(x,digits=max(3, getOption("digits") - 2),
plot(x, dx = 0.2, N = NULL, seed = 1,
within = TRUE, pooling = TRUE, between = FALSE, random = FALSE, ...)
Arguments
x an object of class "plm",
data a data.frame,
subset see lm,
na.action see lm; currently, not fully supported,
effect the effects introduced in the model, one of "individual", "time", or "twoways",
model one of "pooling", "within", "between", "random", "fd", or "ht",
random.method method of estimation for the variance components in the random effects model,
one of "swar" (default), "amemiya", "walhus", or "nerlove",
random.dfcor a numeric vector of length 2 indicating which degree of freedom should be used,
68 plm
inst.method the instrumental variable transformation: one of "bvk", "baltagi", "am", or

"bmc",
index the indexes,
restrict.matrix
a matrix which defines linear restrictions on the coefficients,
restrict.rhs the right hand side vector of the linear restrictions on the coefficients,
digits number of digits for printed output,
width the maximum length of the lines in the printed output,
dx the half–length of the individual lines for the plot method (relative to x range),
N the number of individual to plot,
seed the seed which will lead to individual selection,
within if TRUE, the within model is plotted,
pooling if TRUE, the pooling model is plotted,
between if TRUE, the between model is plotted,
random if TRUE, the random effect model is plotted,
Details
plm is a general function for the estimation of linear panel models. It supports the following es-
timation methods: pooled OLS (model = "pooling"), fixed effects ("within"), random effects
("random"), first–differences ("fd"), and between ("between"). It supports unbalanced panels and
two–way effects (although not with all methods).
For random effects models, four estimators of the transformation parameter are available by setting
random.method to one of "swar" (Swamy and Arora (1972)) (default), "amemiya" (Amemiya
(1971)), "walhus" (Wallace and Hussain (1969)), or "nerlove" (Nerlove (1971)).
For first–difference models, the intercept is maintained (which from a specification viewpoint amounts
to allowing for a trend in the levels model). The user can exclude it from the estimated specification
the usual way by adding "-1" to the model formula.
Instrumental variables estimation is obtained using two–part formulas, the second part indicating
the instrumental variables used. This can be a complete list of instrumental variables or an update
of the first part. If, for example, the model is y ~ x1 + x2 + x3, with x1 and x2 endogenous and
z1 and z2 external instruments, the model can be estimated with:
• formula=y~x1+x2+x3 | x3+z1+z2,
• formula=y~x1+x2+x3 | .-x1-x2+z1+z2.
Balestra and Varadharajan-Krishnakumar’s or Baltagi’s method is used if inst.method="bvk" or

if inst.method="baltagi", respectively.
The Hausman–Taylor estimator is computed if model = "ht".
plm 69
Value
An object of class c("plm","panelmodel").
A "plm" object has the following elements :

vcov the variance–covariance matrix of the coefficients,
formula an object of class 'pFormula' describing the model,
model the model frame as a 'pdata.frame' containing the variables used for estima-
tion: the response is in first column followed by the other variables, the individ-
ual and time indexes are in the ’index’ attribute of model,
ercomp an object of class 'ercomp' providing the estimation of the components of the
errors (for random effects models only),
aliased named logical vector indicating any aliased coefficients which are silently dropped
by plm due to linearly dependent terms (see also detect_lin_dep),
call the call.
It has print, summary and print.summary methods. The summary method creates an object of
class "summary.plm" that extends the object it is run on with information about (inter alia) F statis-
tic and (adjusted) R-squared of model, standard errors, t–values, and p–values of coefficients, (if
supplied) the furnished vcov, see summary.plm for further details.
Author(s)
Yves Croissant
References
Amemiya, T. (1971) The estimation of the variances in a variance–components model, International
Economic Review, 12(1), pp. 1–13.
Balestra, P. and Varadharajan-Krishnakumar, J. (1987) Full information estimations of a system of
simultaneous equations with error components structure, Econometric Theory, 3(2), pp. 223–246.
Baltagi, B.H. (1981) Simultaneous equations with error components, Journal of Econometrics,
17(2), pp. 189–200.
Baltagi, B.H. (2001) Econometric Analysis of Panel Data, 2nd ed., John Wiley and Sons.
Baltagi, B.H. (2013) Econometric Analysis of Panel Data, 5th ed., John Wiley and Sons.
Hausman, J.A. and Taylor W.E. (1981) Panel data and unobservable individual effects, Economet-
rica, 49(6), pp. 1377–1398.
Nerlove, M. (1971) Further evidence on the estimation of dynamic economic relations from a time–
series of cross–sections, Econometrica, 39(2), pp. 359–382.
Swamy, P.A.V.B. and Arora, S.S. (1972) The exact finite sample properties of the estimators of
coefficients in the error components regression models, Econometrica, 40(2), pp. 261–275.
Wallace, T.D. and Hussain, A. (1969) The use of error components models in combining cross
section with time series data, Econometrica, 37(1), pp. 55–72.
70 plm.data
See Also
summary.plm for further details about the associated summary method and the "summary.plm"
object both of which provide some model tests and tests of coefficients. fixef to compute the fixed
effects for "within" models (=fixed effects models).
Examples
summary(zz)
# replicates some results from Baltagi (2013), table 3.1

p <- plm(inv ~ value + capital,
data = Grunfeld, model = "pooling")

data = Grunfeld, model = "within", effect = "twoways")
swar <- plm(inv ~ value + capital,

data = Grunfeld, model = "random", effect = "twoways")
amemiya <- plm(inv ~ value + capital,

data = Grunfeld, model = "random", random.method = "amemiya",
effect = "twoways")
walhus <- plm(inv ~ value + capital,

data = Grunfeld, model = "random", random.method = "walhus",
effect = "twoways")
# summary, summary with a funished vcov, passed as matrix,

# as function, and as function with additional argument
summary(wi)
summary(wi, vcov = vcovHC(wi))
summary(wi, vcov = vcovHC)
summary(wi, vcov = function(x) vcovHC(x, method = "white2"))
plm.data Data Frame Special Format for Panel Data
Description
This function transforms a data frame in a format suitable for using with the estimation functions
of plm.
Usage
plm.data(x, indexes = NULL)
plmtest 71
Arguments
x a data.frame,
indexes a vector (of length one or two) indicating the (individual and time) indexes.
Details
indexes can be:
• a character string which is the name of the individual index variable, in this case a new variable
called “time” containing the time index is added,
• an integer, the number of individuals in the case of balanced panel, in this case two new
variables “time” and “id” containing the individual and the time indexes are added,
• a vector of two character strings which contains the names of the individual and of the time
indexes.
Value
A data.frame.
Author(s)
Yves Croissant
Examples
# There are 595 individuals
Wages <- plm.data(Wages, 595)
# The pdata.frame is called gas
Gasoline <- plm.data(Gasoline, c("country","year"))
summary(Gasoline)

Hedonic <- plm.data(Hedonic, "townid")
plmtest Lagrange Multiplier Tests for Panel Models
Description
Test of individual and/or time effects for panel models.
72 plmtest
Usage
plmtest(x, ...)
plmtest(x, effect = c("individual", "time", "twoways"),
type = c("honda", "bp", "ghm", "kw"), ...)
plmtest(x, data, ..., effect = c("individual", "time", "twoways"),
type = c("honda", "bp", "ghm", "kw"))
Arguments
x an object of class "plm" or a formula of class "formula",
data a data.frame,
effect a character string indicating which effects are tested: individual effects ("individual"),
time effects ("time") or both ("twoways"),
type a character string indicating the test to be performed:
• "honda" (default) for Honda (1985),
• "bp" for Breusch/Pagan (1980),
• "kw" for King/Wu (1997), or
• "ghm" for Gourieroux/Holly/Monfort (1982);
For unbalanced panel data sets, the respective unbalanced version of the tests
are computed,
... further arguments passed to plmtest.
Details
These Lagrange multiplier tests use only the residuals of the pooling model. The first argument of
this function may be either a pooling model of class plm or an object of class formula describing
the model. For inputted within (fixed effects) or random effects models, the corresponding pooling
model is calculated internally first as the tests are based on the residuals of the pooling model.
The bp test for unbalanced panels was derived in Baltagi/Li (1990), the kw test for unbalanced
panels in Baltagi/Chang/Li (1998). The ghm test and the kw test were extended to two-way effects
in Baltagi/Chang/Li (1992).
For a concise overview of all these statistics see Baltagi (2013), Sec. 4.2, pp. 68–76 (for balanced
panels) and Sec. 9.5, pp. 200–203 (for unbalanced panels).
Value
Note
For the King-Wu statistics ("kw"), the oneway statistics ("individual" and "time") coincide with
the respective Honda statistics ("honda"); twoway statistics of "kw" and "honda" differ.
plmtest 73
Author(s)
Yves Croissant (initial implementation), Kevin Tappe (generalization to unbalanced panels)
References
Baltagi, B. H. (2013) Econometric Analysis of Panel Data, 5th edition, Sec. 4.2, pp. 68–76 and
Sec. 9.5, pp. 200–203.
Baltagi, B. H./Li, Q. (1990) A lagrange multiplier test for the error components model with incom-
plete panels, Econometric Reviews, 9(1), pp. 103–107.
Baltagi B. H./Chang, Y./Li, Q. (1992) Monte Carlo results on several new and existing tests for the
error component model, Journal of Econometrics, 54(1-3), pp. 95–120.
Baltagi B. H./Chang, Y./Li, Q. (1998) Testing for random effects using unbalanced panel data,
Advances in Econometrics, 13, pp. 1–20.
Breusch, T. S./Pagan, A. R. (1980) The Lagrange multiplier test and its applications to model spec-
ification in econometrics, Review of Economic Studies, 47(1), pp. 239–253.
Gourieroux, C./Holly, A./Monfort, A. (1982) Likelihood ratio test, Wald test, and Kuhn–Tucker test
in linear models with inequality constraints on the regression parameters, Econometrica, 50(1), pp.
63–80.
Honda, Y. (1985) Testing the error components model with non–normal disturbances, Review of
Economic Studies, 52(4), pp. 681–690.
King, M. L./Wu, P. X. (1997) Locally optimal one–sided tests for multiparameter hypotheses,
Econometric Reviews, 16(2), pp. 131–156.
See Also
pFtest for individual and/or time effects tests based on the within model.
Examples
g <- plm(inv ~ value + capital, data = Grunfeld, model = "pooling")
plmtest(g)
plmtest(g, effect="time")
plmtest(inv ~ value + capital, data = Grunfeld, type = "honda")
plmtest(inv ~ value + capital, data = Grunfeld, type = "bp")
plmtest(inv ~ value + capital, data = Grunfeld, type = "bp", effect = "twoways")
plmtest(inv ~ value + capital, data = Grunfeld, type = "ghm", effect = "twoways")
plmtest(inv ~ value + capital, data = Grunfeld, type = "kw", effect = "twoways")
Grunfeld_unbal <- Grunfeld[1:(nrow(Grunfeld)-1), ] # create an unbalanced panel data set

g_unbal <- plm(inv ~ value + capital, data = Grunfeld_unbal, model = "pooling")
plmtest(g_unbal) # unbalanced version of test is indicated in output
74 pmg
pmg Mean Groups (MG), Demeaned MG and CCE MG estimators
Description
Mean Groups (MG), Demeaned MG (DMG) and Common Correlated Effects MG (CCEMG) esti-
mators for heterogeneous panel models, possibly with common factors (CCEMG)
Usage
pmg(formula, data, subset, na.action,
model = c("mg", "cmg", "dmg"),
index = NULL, trend = FALSE, ...)
## S3 method for class 'pmg'
## S3 method for class 'summary.pmg'
print(x,digits = max(3, getOption("digits") -
Arguments
object, x an object of class pmg,
data a data.frame,
subset see lm,
na.action see lm,
model one of c("mg", "cmg", "dmg"),
trend logical specifying whether an individual-specific trend has to be included,
digits digits,
Details
pmg is a function for the estimation of linear panel models with heterogeneous coefficients by the
Mean Groups estimator. model="mg" specifies the standard Mean Groups estimator, based on
the average of individual time series regressions. If model="dmg" the data are demeaned cross-
sectionally, which is believed to reduce the influence of common factors (and is akin to what is
done in homogeneous panels when model="within" and effect="time". Lastly, if model="cmg"
then the CCEMG estimator is employed: this latter is consistent under the hypothesis of unob-
served common factors and idiosyncratic factor loadings; it works by augmenting the model by
cross-sectional averages of the dependent variable and regressors in order to account for the com-
mon factors, and adding individual intercepts and possibly trends.
pmg 75
Value
An object of class c("pmg", "panelmodel") containing:

call the call,
sigma always NULL, sigma is here only for compatibility reasons (to allow using the
same summary and print methods as pggls),
indcoef the matrix of individual coefficients from separate time series regressions.
Author(s)
Giovanni Millo
References
M. Hashem Pesaran (2006), Estimation and Inference in Large Heterogeneous Panels with a Multi-
factor Error Structure, Econometrica, 74(4), pp. 967–1012.
Examples

## Mean Groups estimator
mgmod <- pmg(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp, data = Produc)
summary(mgmod)
## demeaned Mean Groups

dmgmod <- pmg(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp,
data = Produc, model="dmg")
summary(dmgmod)
## Common Correlated Effects Mean Groups

ccemgmod <- pmg(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp,
data = Produc, model="cmg")
summary(ccemgmod)
76 pmodel.response
pmodel.response A function to extract the model.response
Description
pmodel.response has several methods to conveniently extract the response of several objects.
Usage
pmodel.response(object, ...)
pmodel.response(object,
lhs = NULL,
theta = NULL, ...)
pmodel.response(object, data,
lhs = NULL,
theta = NULL, ...)
pmodel.response(object, ...)
Arguments
object an object of class "plm", or a formula of class "pFormula",
data a pdata.frame, which is a model frame (if not, it is tried to construct the model
frame from the data, see Details),
effect the effects introduced in the model, one of "individual", "time" or "twoways",
model one of "pooling", "within", "between", "random", "fd" and "ht",
theta the parameter for the transformation if model = "random",
lhs inherited from package Formula (see there),
Details
The model response is extracted from a pdata.frame (where the response must reside in the first
column; this is the case for a model frame), a pFormula + data or a plm object, and the transfor-
mation specified by effect and model is applied to it.
Constructing the model frame first ensures proper NA handling and the response being placed in
the first column, see also Examples for usage.
pooltest 77
Value
A numeric vector.
Author(s)
Yves Croissant
See Also
plm’s model.matrix for (transformed) model matrix and the corresponding model.frame method
to construct a model frame.
Examples
data(Grunfeld)

# construct (transformed) response of the within model

resp <- pmodel.response(pform, data = mf, model = "within")
# retrieve (transformed) response directly from model frame

resp_mf <- pmodel.response(mf, model = "within")
# retrieve (transformed) response from a plm object, i.e. an estimated model

fe_model <- plm(pform, data = pGrunfeld, model = "within")
pmodel.response(fe_model)
# same as constructed before

all.equal(resp, pmodel.response(fe_model), check.attributes = FALSE) # TRUE
pooltest Test of Poolability
Description
A Chow test for the poolability of the data.
Usage
pooltest(x, ...)
pooltest(x, z, ...)
pooltest(x, data, ...)
78 Produc
Arguments
x an object of class "plm",
z an object of class "pvcm" obtained with model="within",
data a data.frame,
... further arguments passed to plm.
Details
pooltest is an F test of stability (or Chow test) for the coefficients of a panel model. The estimated
plm object should be a "pooling" model or a "within" model (the default); intercepts are assumed
to be identical in the first case and different in the second case.
Value
Author(s)
Yves Croissant
Examples
form <- lgaspcar ~ lincomep + lrpmg + lcarpcap
gasw <- plm(form, data = Gasoline, model = "within")
gasp <- plm(form, data = Gasoline, model = "pooling")
gasnp <- pvcm(form, data = Gasoline, model = "within")
pooltest(gasw, gasnp)
pooltest(gasp, gasnp)
pooltest(form, data = Gasoline, effect = "individual", model = "within")

pooltest(form, data = Gasoline, effect = "individual", model = "pooling")
Produc US States Production
Description
Usage
data(Produc)
pseries 79
Format
state the state

year the year
region the region
pcap public capital stock
hwy highway and streets
water water and sewer facilities
util other public buildings and structures
pc private capital stock
gsp gross state product
emp labor input measured by the employment in non–agricultural payrolls
unemp state unemployment rate
Source
resourceId=13452
References
Baltagi, B. H. and N. Pinnoi (1995) “Public capital stock and state productivity growth: further
evidence”, Empirical Economics, 20(2), pp. 351–359.
Munnell, A.H. (1990) “Why has productivity growth declined? Productivity and public invest-
ment”, New England Economic Review, pp. 3–22.
pseries panel series
Description
A class for panel series for which several useful computations and data transformations are avail-
able.
80 pseries
Usage
between(x, ...)
Between(x, ...)
Within(x, ...)
between(x, effect = c("individual", "time"), ...)
Between(x, effect = c("individual", "time"), ...)
Within(x, effect = c("individual", "time"), ...)
## S3 method for class 'summary.pseries'
print(x, ...)
as.matrix(x, idbyrow = TRUE, ...)
Arguments
x, object a pseries or a summary.pseries object,
effect character string indicating the "individual" or "time" effect,
idbyrow if TRUE in the as.matrix method, the lines of the matrix are the individuals,
... further arguments, e. g. na.rm = TRUE for transformation functions like
beetween, see Details and Examples.
Details
The functions between, Between, and Within perform specific data transformations, i. e. the
between and within transformation.
between returns a vector containing the individual means (over time) with the length of the vector
equal to the number of individuals (if effect = "individual" (default); if effect = "time", it
returns the time means (over individuals)). Between duplicates the values and returns a vector which
length is the number of total observations. Within returns a vector containing the values in deviation
from the individual means (if effect = "individual", from time means if effect = "time"),
the so called demeaned data.
For between, Between, and Within in presence of NA values it can be useful to supply na.rm = TRUE
as an additional argument to keep as many observations as possible in the resulting transformation,
see also Examples.
Value
All these functions return an object of class pseries, except:
between, which returns a numeric vector, as.matrix, which returns a matrix.
Author(s)
Yves Croissant
punbalancedness 81
See Also
For more functions on class ’pseries’ see lag, lead, diff for lagging values, leading values (nega-
tive lags) and differencing.
Examples
# First, create a pdata.frame
Em <- pdata.frame(EmplUK)
# Then extract a series, which becomes additionally a pseries

z <- Em$output
class(z)
# obtain the matrix representation

as.matrix(z)
# compute the between and within transformations

between(z)
Within(z)
# Between replicates the values for each time observation

Between(z)
# between, Between, and Within transformations on other dimension

between(z, effect = "time")
Between(z, effect = "time")
Within(z, effect = "time")
# NA treatment for between, Between, and Within

z2 <- z
z2[length(z2)] <- NA # set last value to NA
between(z2, na.rm = TRUE) # non-NA value for last individual
Between(z2, na.rm = TRUE) # only the NA observation is lost
Within(z2, na.rm = TRUE) # only the NA observation is lost
sum(is.na(Between(z2))) # 9 observations lost due to one NA value

sum(is.na(Between(z2, na.rm = TRUE))) # only the NA observation is lost
sum(is.na(Within(z2))) # 9 observations lost due to one NA value
sum(is.na(Within(z2, na.rm = TRUE))) # only the NA observation is lost
punbalancedness Measures for Unbalancedness of Panel Data
Description
This function reports two unbalancedness measures for panel data as defined in Ahrens/Pincus
(1981).
82 punbalancedness
Usage
punbalancedness(x, ...)
punbalancedness(x, index = NULL, ...)
Arguments
x a panelmodel, a data.frame, or a pdata.frame object,
index only relevant for data.frame interface, for details see pdata.frame,
Details
punbalancedness returns two measures for the unbalancedness of a panel data set (called "gamma"
(γ) and "nu" (ν)).
If the panel data are balanced, both measures equal 1. The more "unbalanced" the panel data, the
lower the measures (but > 0). The upper and lower bounds as given in Ahrens/Pincus (1981) are:
0 < γ, ν ≤ 1, and for ν more precisely n1 < ν ≤ 1, with n being the number of individuals (as in
pdim(x)$nT$n).
An application of the first measure ("gamma") is found in e. g. Baltagi/Song/Jung (2002), pp.
488-491, and Baltagi/Chang (1994), pp. 78–87, where it is used to measure the unbalancedness of
various unbalanced data sets used for Monte Carlo simulation studies.
There exists also an extension of unbalancedness measures to nested panel structures as developed
in Baltagi/Song/Jung (2001), p. 368, but these are not implemented in punbalancedness as of now.
punbalancedness uses output of pdim to calculate the unbalancedness measures, so inputs to
punbalancedness can be whatever pdim works on. pdim returns a logical whether a panel data
set is balanced or not and detailed information about the number of individuals and time observa-
tions (see pdim).
Value
A numeric vector containing two entries (in this order):
gamma unbalancedness measure "gamma" (γ) (as called by the Greek letter in Ahrens/Pincus
(1981), p. 228),
nu unbalancedness measure "nu" (ν) (as called by the Greek letter in Ahrens/Pincus
(1981), p. 228).
Note
Calling punbalancedness on an estimated panelmodel object and on the corresponding (p)data.frame
used for this estimation does not necessarily yield the same result (true also for pdim). When
called on an estimated panelmodel, the number of observations (individual, time) actually used
punbalancedness 83
for model estimation are taken into account. When called on a (p)data.frame, the rows in the
(p)data.frame are considered, disregarding any NA values in the dependent or independent vari-
able(s) which would be dropped during model estimation.
Author(s)
Kevin Tappe
References
Ahrens, H.; Pincus, R. (1981), “On two measures of unbalancedness in a one-way model and their
relation to efficiency”, Biometrical Journal, 23(3), pp. 227–235.
Baltagi, Badi H.; Chang, Young-Jae (1994), “Incomplete panels: A comparative study of alternative
estimators for the unbalanced one-way error component regression model”, Journal of Economet-
rics, 62(2), pp. 67–89.
Baltagi, Badi H.; Song, Seuck Heun; Jung, Byoung Cheol (2001), “The unbalanced nested error
component regression model”, Journal of Econometrics, 101(2), pp. 357–381.
Baltagi, Badi H.; Song, Seuck H.; Jung, Byoung C. (2002), “A comparative study of alternative
estimators for the unbalanced two-way error component regression model”, Econometrics Journal,
5(2), pp. 480–493.
See Also
nobs, pdim, pdata.frame
Examples
# Grunfeld is a balanced panel, Hedonic is an unbalanced panel
data(list=c("Grunfeld", "Hedonic"), package="plm")
# Grunfeld has individual and time index in first two columns

punbalancedness(Grunfeld) # c(1,1) indicates balanced panel
pdim(Grunfeld)$balanced # TRUE
# Hedonic has individual index in column "townid" (in last column)

punbalancedness(Hedonic, index="townid") # c(0.472, 0.519)
pdim(Hedonic, index="townid")$balanced # FALSE
# punbalancedness on estimated models

plm_mod_pool <- plm(inv ~ value + capital, data = Grunfeld)
punbalancedness(plm_mod_pool)
plm_mod_fe <- plm(inv ~ value + capital, data = Grunfeld[1:99, ], model = "within")

punbalancedness(plm_mod_fe)
# replicate results for panel data design no. 1 in Ahrens/Pincus (1981), p. 234
ind_d1 <- c(1,1,1,2,2,2,3,3,3,3,3,4,4,4,4,4,4,4,5,5,5,5,5,5,5)
time_d1 <- c(1,2,3,1,2,3,1,2,3,4,5,1,2,3,4,5,6,7,1,2,3,4,5,6,7)
df_d1 <- data.frame(individual = ind_d1, time = time_d1)
punbalancedness(df_d1) # c(0.868, 0.887)
84 purtest
purtest Unit root tests for panel data
Description
purtest implements several testing procedures that have been proposed to test unit root hypotheses
with panel data.
Usage
purtest(object, data = NULL, index = NULL,
test= c("levinlin", "ips", "madwu", "hadri"),
exo = c("none", "intercept", "trend"),
lags = c("SIC", "AIC", "Hall"), pmax = 10, Hcons = TRUE,
q = NULL, dfcor = FALSE, fixedT = TRUE, ...)
## S3 method for class 'purtest'
print(x, ...)
## S3 method for class 'purtest'
## S3 method for class 'summary.purtest'
print(x, ...)
Arguments
object, x Either a 'data.frame' or a matrix containing the time series, a 'pseries' ob-
ject, a formula, or the name of a column of a 'data.frame', or a 'pdata.frame'
on which the test has to be computed; a 'purtest' object for the print and sum-
mary methods,
data a 'data.frame' or a 'pdata.frame' object,
index the indexes,
test the test to be computed: one of "levinlin" for Levin, Lin and Chu (2002),
"ips" for Im, Pesaran and Shin (2003), "madwu" for Maddala and Wu (1999),
or "hadri" for Hadri (2000),
exo the exogenous variables to introduce in the augmented Dickey–Fuller regres-
sions, one of: no exogenous variables ("none"), individual intercepts ("intercept"),
or individual intercepts and trends ("trend"),
lags the number of lags to be used for the augmented Dickey-Fuller regressions:
either an integer (the number of lags for all time series), a vector of integers
(one for each time series), or a character string for an automatic computation
of the number of lags, based on either the AIC ("AIC"), the SIC ("SIC"), or on
Hall’s method ("Hall"),
pmax maximum number of lags,
Hcons logical, indicating whether the heteroscedasticity-consistent test of Hadri should
be computed,
q the bandwidth for the estimation of the long-run variance,
purtest 85
dfcor logical, indicating whether the standard deviation of the regressions is to be

computed using a degrees-of-freedom correction,
fixedT logical, indicating whether the different ADF regressions are to be computed
using the same number of observations,
Details
All these tests except "hadri" are based on the estimation of augmented Dickey-Fuller regressions
for each time series. A statistic is then computed using the t-statistic associated with the lagged
variable.
The kind of test to be computed can be specified in several ways:
A formula/data interface (if data is a data.frame, an additional index argument should be spec-
ified); the formula should be of the form: y~0, y~1, or y~trend for a test with no exogenous
variables, with an intercept, or with individual intercepts and time trend, respectively.
A data.frame, a matrix, a pseries: in this case, the exogenous variables are specified using the
exo argument.
The Hadri statistic is the cross-sectional average of the individual KPSS statistics, standardized by
their asymptotic mean and standard deviation.
Value
An object of class 'purtest': a list with the elements 'statistic' (a 'htest' object), 'call',
'args', 'idres' (containing results from the individual regressions), and 'adjval' (containing
the simulated means and variances needed to compute the statistic).
Author(s)
Yves Croissant
References
Hadri K. (2000). “Testing for Stationarity in Heterogeneous Panel Data”, The Econometrics Jour-
nal, 3(2), pp. 148–161.
Im K.S., Pesaran M.H. and Shin Y. (2003). “Testing for Unit Roots in Heterogeneous Panels”,
Journal of Econometrics, 115(1), pp. 53–74.
Levin A., Lin C.F. and Chu C.S.J. (2002). “Unit Root Tests in Panel Data: Asymptotic and Finite
Sample Properties”, Journal of Econometrics, 108(1), pp. 1–24.
Maddala G.S. and Wu S. (1999). “A Comparative Study of Unit Root Tests with Panel Data and a
New Simple Test”, Oxford Bulletin of Economics and Statistics, 61, Supplement 1, pp. 631–652.
Examples
y <- data.frame(split(Grunfeld$inv, Grunfeld$firm))
purtest(y, pmax = 4, exo = "intercept", test = "madwu")

86 pvar
## same via formula interface

purtest(inv ~ 1, data = Grunfeld, index = "firm", pmax = 4, test = "madwu")
pvar Check for Cross-Sectional and Time Variation
Description
This function checks for each variable of a panel if it varies cross-sectionally and over time.
Usage
pvar(x, ...)
pvar(x, index = NULL, ...)
pvar(x, ...)
## S3 method for class 'matrix'
pvar(x, index = NULL, ...)
## S3 method for class 'pvar'
print(x, ...)
Arguments
x a (p)data.frame or a matrix,
index see pdata.frame,
Details
For (p)data.frame and matrix interface: All-NA columns are removed prior to calculation of varia-
tion due to coercing to pdata.frame first.
Value
An object of class pvar containing the following elements:
id.variation a logical vector with TRUE values if the variable has individual variation, FALSE
if not,
time.variation a logical vector with TRUE values if the variable has time variation, FALSE if not,
id.variation_anyNA
a logical vector with TRUE values if the variable has at least one individual-time
combination with all NA values in the individual dimension for at least one time
period, FALSE if not,
time.variation_anyNA
a logical vector with TRUE values if the variable has at least one individual-time
combination with all NA values in the time dimension for at least one individual,
FALSE if not.
pvcm 87
Note
pvar can be time consuming for “big” panels.
Author(s)
Yves Croissant
See Also
pdim to check the dimensions of a ’pdata.frame’ (and other objects),
Examples
# and are the first two variables
pvar(Gasoline)
# Hedonic is an unbalanced panel, townid is the individual index;

# the drop.index argument is passed to pdata.frame
pvar(Hedonic, "townid", drop.index = TRUE)
# same using pdata.frame

Hed <- pdata.frame(Hedonic, "townid", drop.index = TRUE)
pvar(Hed)
# Gasoline with pvar's matrix interface

Gasoline_mat <- as.matrix(Gasoline)
pvar(Gasoline_mat)
pvar(Gasoline_mat, index=c("country", "year"))
pvcm Variable Coefficients Models for Panel Data
Description
Estimators for random and fixed effects models with variable coefficients.
Usage
pvcm(formula, data, subset, na.action, effect = c("individual","time"),
model = c("within","random"), index = NULL, ...)
## S3 method for class 'pvcm'
## S3 method for class 'summary.pvcm'
print(x, digits = max(3, getOption("digits") -2),
88 pvcm
Arguments
object, x an object of class "pvcm",
data a data.frame,
subset see lm,
na.action see lm,
effect the effects introduced in the model: one of "individual", "time",
model one of "within", "random",
digits digits,
Details
pvcm estimates variable coefficients models. Time or individual effects are introduced, respectively,
if effect="time" or effect="individual" (the default value).
Coefficients are assumed to be fixed if model="within" and random if model="random". In the
first case, a different model is estimated for each individual (or time period). In the second case, the
Swamy (1970) model is estimated. It is a generalized least squares model which uses the results of
the previous model.
Value
An object of class c("pvcm","panelmodel"), which has the following elements :
coefficients the vector (or the list for fixed effects) of coefficients,
call the call,
Delta the estimation of the covariance matrix of the coefficients (random effect models
only),
std.error the standard errors for all the coefficients for each individual (within models
only),
pvcm objects have print, summary and print.summary methods.
Author(s)
Yves Croissant
pwaldtest 89
References
Swamy, P.A.V.B. (1970). Efficient Inference in a Random Coefficient Regression Model, Econo-
metrica, 38(2), pp.311–323.
Examples
zw <- pvcm(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp, data = Produc, model = "within")
zr <- pvcm(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp, data = Produc, model = "random")
pwaldtest Wald-style Chi-square Test and F Test
Description
Wald-style chi-square test and F test of slope coefficients being zero jointly, including robust ver-
sions of the tests.
Usage
pwaldtest(x, ...)
pwaldtest(x, test = c("Chisq", "F"), vcov = NULL,
df2adj = (test == "F" && !is.null(vcov) && missing(.df2)),
.df1, .df2, ...)
Arguments
x an estimated model of which the coefficients should be tested (usually of class
"plm"),
test a character, indicating the test to be performed, may be either "Chisq" or "F"
for the Wald-style Chi-square test or F test, respectively,
vcov NULL by default; a matrix giving a variance–covariance matrix or a function
which computes such; if supplied (non NULL), the test is carried out using the
variance–covariance matrix indicated resulting in a robust test,
df2adj logical, only relevant for test = "F", indicating whether the adjustment for
clustered standard errors for the second degrees of freedom parameter should
be performed (see Details, also for further requirements regarding the variance–
covariance matrix in vcov for the adjustment to be performed),
.df1 a numeric, used if one wants to overwrite the first degrees of freedom parameter
in the performed test (usually not used),
.df2 a numeric, used if one wants to overwrite the second degrees of freedom param-
eter for the F test (usually not used),
... further arguments (currently none).
90 pwaldtest
Details
pwaldtest can be used stand–alone with a plm object. It is also used in summary.plm to produce
the F statistic.
pwaldtest performs the test if the slope coefficients of a panel regression are jointly zero. It does
not perform general purpose Wald-style tests (for those, see waldtest (from package lmtest) or
linearHypothesis (from car)).
If a user specified variance-covariance matrix/function is given in argument vcov, the robust version
of the tests are carried out. In that case, if the F test is requested (test = "F") and no overwriting of
the second degrees of freedom parameter is given (by supplying argument (.df2)), the adjustment
of the second degrees of freedom parameter is performed by default. The second degrees of freedom
parameter is adjusted to be the number of unique elements of the clustered variable - 1, e. g. the
number of individuals - 1.
The adjustment requires the vcov to carry an attribute called "cluster" with a known clustering de-
scribed as a character (for now this could be either the character "group" or "time"). The vcovXX
functions of the package plm provide such an attribute for their returned variance–covariance ma-
trices. No adjustment is done for unknown descriptions in the character of the attribute "cluster" or
when the attribute "cluster" is not present. For the adjustment, see e. g. Cameron/Miller (2015),
section VII; Andress/Golsch/Schmidt (2013), pp. 126, footnote 4.
Value
Author(s)
Yves Croissant (initial implementation) and Kevin Tappe (extensions: vcov argument and df2 ad-
justment)
References
Press, Sec. 4.2.3, pp. 60–62.
Andress, H.-J./Golsch, K./Schmidt, A. (2013), Applied Panel Data Analysis for Economic and
Social Surveys, Springer, Heidelberg et al.
Cameron, A. C./Miller, D. L. (2015), "A Practitioner’s Guide to Cluster-Robust Inference", Journal
of Human Resources, 2015, Vol. 50, No. 2, pp. 317–373; see also the supplements under http:
//cameron.econ.ucdavis.edu/research/papers.html.
See Also
vcovHC for an example of the vcovXX functions, a robust estimation for the variance–covariance
matrix; summary.plm
Examples
mod_fe <- plm(inv ~ value + capital, data = Grunfeld, model = "within")
mod_re <- plm(inv ~ value + capital, data = Grunfeld, model = "random")
pwartest 91
pwaldtest(mod_fe, test = "F")

pwaldtest(mod_re, test = "Chisq")
# with robust vcov

pwaldtest(mod_fe, vcov = vcovHC(mod_fe))
pwaldtest(mod_fe, vcov = function(x) vcovHC(x, type = "HC3"))
pwaldtest(mod_fe, vcov = vcovHC(mod_fe), df2adj = FALSE) # w/o df2 adjustment
# example without attribute "cluster" in the vcov

vcov_mat <- vcovHC(mod_fe)
attr(vcov_mat, "cluster") <- NULL # remove attribute
pwaldtest(mod_fe, vcov = vcov_mat) # no df2 adjustment performed
pwartest Wooldridge Test for AR(1) Errors in FE Panel Models
Description
Test of serial correlation for (the idiosyncratic component of) the errors in fixed–effects panel mod-
els.
Usage
pwartest(x, ...)
pwartest(x, ...)
pwartest(x, data, ...)
Arguments
x an object of class formula or of class panelmodel,
data a data.frame,
... further arguments to be passed on to linearHypothesis or to vcovHC.
Details
As Wooldridge (2003/2010, Sec. 10.5.4) observes, under the null of no serial correlation in the
errors, the residuals of a FE model must be negatively serially correlated, with cor(ûit , ûis ) =
−1/(T − 1) for each t, s. He suggests basing a test for this null hypothesis on a pooled regression
of FE residuals on their first lag: ûi,t = α+δ ûi,t−1 +ηi,t . Rejecting the restriction δ = −1/(T −1)
makes us conclude against the original null of no serial correlation.
pwartest estimates the within model and retrieves residuals, then estimates an AR(1) pooling
model on them. The test statistic is obtained by applying linearHypothesis() to the latter
model to test the above restriction on δ, setting the covariance matrix to vcovHC with the option
method="arellano" to control for serial correlation.
92 pwfdtest
Unlike the pbgtest and pdwtest, this test does not rely on large–T asymptotics and has therefore
good properties in “short” panels. Furthermore, it is robust to general heteroskedasticity.
Value
Author(s)
Giovanni Millo
References
Wooldridge, J.M. (2002) Econometric Analysis of Cross-Section and Panel Data, MIT Press, Sec.
10.5.4, pp. 274–276.
Wooldridge, J.M. (2010) Econometric analysis of cross-section and panel data, 2nd ed., MIT Press,
Sec. 10.5.4, pp. 310–312.
See Also
pwfdtest, pdwtest, pbgtest, pbltest, pbsytest.
Examples
pwartest(log(emp) ~ log(wage) + log(capital), data = EmplUK)
pwfdtest Wooldridge first–difference–based test for AR(1) errors in levels or

first–differenced panel models
Description
First–differencing–based test of serial correlation for (the idiosyncratic component of) the errors in
either levels or first–differenced panel models.
Usage
pwfdtest(x, ...)
pwfdtest(x, ..., h0 = c("fd", "fe"))
pwfdtest(x, data, ..., h0 = c("fd", "fe"))
pwfdtest 93
Arguments
x an object of class formula or a "fd"-model (plm object),
data a data.frame,
h0 the null hypothesis: one of "fd", "fe",
... further arguments to be passed on to linearHypothesis or to vcovHC.
Details
As Wooldridge (2003/2010, Sec. 10.6.3) observes, if the idiosyncratic errors in the model in levels
are uncorrelated (which we label hypothesis "fe"), then the errors of the model in first differences
(FD) must be serially correlated with cor(êit , êis ) = −0.5 for each t, s. If on the contrary the
levels model’s errors are a random walk, then there must be no serial correlation in the FD errors
(hypothesis "fd"). Both the fixed effects (FE) and the first–differenced (FD) estimators remain
consistent under either assumption, but the relative efficiency changes: FE is more efficient under
"fe", FD under "fd".
Wooldridge (ibid.) suggests basing a test for either hypothesis on a pooled regression of FD residu-
als on their first lag: êi,t = α+ρêi,t−1 +ηi,t . Rejecting the restriction ρ = −0.5 makes us conclude
against the null of no serial correlation in errors of the levels equation ("fe"). The null hypothesis
of no serial correlation in differenced errors ("fd") is tested in a similar way, but based on the zero
restriction on ρ (ρ = 0). Rejecting "fe" favours the use of the first–differences estimator and the
contrary, although it is possible that both be rejected.
pwfdtest estimates the fd model (or takes an fd model as input for the panelmodel interface)
and retrieves its residuals, then estimates an AR(1) pooling model on them. The test statistic is
obtained by applying linearHypothesis() from package car to the latter model to test the relevant
restriction on ρ, setting the covariance matrix to vcovHC with the option method="arellano" to
control for serial correlation.
Unlike the pbgtest and pdwtest, this test does not rely on large–T asymptotics and has therefore
good properties in ”short” panels. Furthermore, it is robust to general heteroskedasticity. The
"fe" version can be used to test for error autocorrelation regardless of whether the maintained
specification has fixed or random effects (see Drukker (2003)).
Value
Author(s)
Giovanni Millo
References
Drukker, D.M. (2003) Testing for serial correlation in linear panel–data models, The Stata Journal,
3(2), pp. 168–177.
Wooldridge, J.M. (2003) Econometric analysis of cross–section and panel data, MIT Press, Sec.
10.6.3, pp. 282–283.
Wooldridge, J.M. (2010) Econometric analysis of cross–section and panel data, 2nd ed., MIT Press,
Sec. 10.6.3, pp. 319–320.
94 pwtest
See Also
pdwtest, pbgtest, pwartest,
Examples
data(EmplUK)
pwfdtest(log(emp) ~ log(wage) + log(capital), data = EmplUK)
pwfdtest(log(emp) ~ log(wage) + log(capital), data = EmplUK, h0 = "fe")
# pass argument 'type' to vcovHC used in test

pwfdtest(log(emp) ~ log(wage) + log(capital), data = EmplUK, type = "HC3", h0 = "fe")
# same with panelmodel interface

mod <- plm(log(emp) ~ log(wage) + log(capital), data = EmplUK, model = "fd")
pwfdtest(mod)
pwfdtest(mod, h0 = "fe")
pwfdtest(mod, type = "HC3", h0 = "fe")
pwtest Wooldridge’s Test for Unobserved Effects in Panel Models
Description
Semi-parametric test for the presence of (individual or time) unobserved effects in panel models.
Usage
pwtest(x, ...)
pwtest(x, data, effect = c("individual", "time"), ...)
pwtest(x, effect = c("individual", "time"), ...)
Arguments
x an object of class "formula", or an estimated model of class panelmodel,

effect the effect to be tested for, one of "individual" (default) or "time",
data a data.frame,
... further arguments passed to plm.
pwtest 95
Details
This semi-parametric test checks the null hypothesis of zero correlation between errors of the same
group. Therefore, it has power both against individual effects and, more generally, any kind of serial
correlation.
The test relies on large-N asymptotics. It is valid under error heteroskedasticity and departures from
normality.
The above is valid if effect="individual", which is the most likely usage. If effect="time",
symmetrically, the test relies on large-T asymptotics and has power against time effects and, more
generally, against cross-sectional correlation.
If the panelmodel interface is used, the inputted model must be a pooling model.
Value
Author(s)
Giovanni Millo
References
Wooldridge, J.M. (2002) Econometric Analysis of Cross-Section and Panel Data, MIT Press, Sec.
10.4.4., pp. 264–265.
Press, Sec. 10.4.4, pp. 299–230.
See Also
pbltest, pbgtest, pdwtest, pbsytest, pwartest, pwfdtest for tests for serial correlation in
panel models. plmtest for tests for random effects.
Examples
## formula interface
pwtest(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp, data = Produc)
pwtest(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp, data = Produc, effect = "time")
## panelmodel interface
# first, estimate a pooling model, than compute test statistics
form <- formula(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp)
pool_prodc <- plm(form, data = Produc, model = "pooling")
pwtest(pool_prodc) # == effect="individual"
pwtest(pool_prodc, effect="time")
96 r.squared
r.squared R squared and adjusted R squared for panel models
Description
This function computes R squared or adjusted R squared for plm objects. It allows to define on
which transformation of the data the (adjusted) R squared is to be computed and which method for
calculation is used.
Usage
r.squared(object, model = NULL, type = c("cor", "rss", "ess"), dfcor = FALSE)
Arguments
object an object of class "plm",

model on which transformation of the data the R-squared is to be computed. If NULL,
the transformation used to estimate the model is also used for the computation
of R squared,
type indicates method which is used to compute R squared. One of
"rss" (residual sum of squares),
"ess" (explained sum of squares), or
"cor" (coefficient of correlation between the fitted values and the response),
dfcor if TRUE, the adjusted R squared is computed.
Value
A numerical value. The R squared or adjusted R squared of the model estimated on the transformed
data, e. g. for the within model the so called "within R squared".
See Also
plm for estimation of various models; summary.plm which makes use of r.squared.
Examples
p <- plm(inv ~ value + capital, data = Grunfeld, model = "pooling")
r.squared(p)
r.squared(p, dfcor = TRUE)
RiceFarms 97
RiceFarms Production of Rice in India
Description
a panel of 171 observations
observation : farms
country : Indonesia
Usage
data(RiceFarms)
Format
A dataframe containing :
id the farm identifier
size the total area cultivated with rice, measured in hectares
status land status, on of 'owner' (non sharecroppers, owner operators or leaseholders or both),
'share' (sharecroppers), 'mixed' (mixed of the two previous status)
varieties one of 'trad' (traditional varieties), 'high' (high yielding varieties) and 'mixed' (mixed
varieties)
bimas bIMAS is an intensification program; one of 'no' (non-bimas farmer), 'yes' (bimas farmer)
or 'mixed' (part but not all of farmer’s land was registered to be in the bimas program)
seed seed in kilogram
urea urea in kilogram
phosphate phosphate in kilogram
pesticide pesticide cost in Rupiah
pseed price of seed in Rupiah per kg
purea price of urea in Rupiah per kg
pphosph price of phosphate in Rupiah per kg
hiredlabor hired labor in hours
famlabor family labor in hours
totlabor total labor (excluding harvest labor)
wage labor wage in Rupiah per hour
goutput gross output of rice in kg
noutput net output, gross output minus harvesting cost (paid in terms of rice)
price price of rough rice in Rupiah per kg
region one of 'wargabinangun', 'langan', 'gunungwangi', 'malausma', 'sukaambit', 'ciwangi'
98 sargan
Source
Qu Feng and William C. Horrace, (2012) “Alternative Measures of Technical Efficiency: Skew,
Bias and Scale”, Journal of Applied Econometrics, 27(2), pp. 253–268.
sargan Hansen–Sargan Test of Overidentifying Restrictions
Description
A test of overidentifying restrictions for models estimated by GMM.
Usage
sargan(object, weights = c("twosteps", "onestep"))
Arguments
object an object of class "pgmm",
weights the weighting matrix to be used for the computation of the test.
Details
The Hansen–Sargan test calculates the quadratic form of the moment restrictions that is minimized
while computing the GMM estimator. It follows asymptotically a chi-square distribution with num-
ber of degrees of freedom equal to the difference between the number of moment conditions and
the number of coefficients.
Value
Author(s)
Yves Croissant
References
Hansen, L.P. (1982), Large Sample Properties of Generalized Methods of Moments Estimators,
Econometrica, 50(4), pp. 1029–1054.
Sargan, J.D. (1958), The Estimation of Economic Relationships using Instrumental Variables, Econo-
metrica, 26(3), pp. 393–415.
See Also
pgmm
Snmesp 99
Examples
ar <- pgmm(log(emp) ~ lag(log(emp), 1:2) + lag(log(wage), 0:1) +
lag(log(capital), 0:2) + lag(log(output), 0:2) | lag(log(emp), 2:99),
sargan(ar)
Snmesp Employment and Wages in Spain
Description
total number of observations: 5904
observation: firms
country: Spain
Usage
data(Snmesp)
Format
A data frame containing:
firm firm index

year year
n log of employment
w log of wages
y log of real output
i log of intermediate inputs
k log of real capital stock
f real cash flow
Source
Journal of Business Economics and Statistics data archive:
http://amstat.tandfonline.com/loi/ubes20/.
References
Alonso–Borrego, C. and Arellano, M. (1999). Symmetrically Normalized Instrumental-Variable
Estimation Using Panel Data, Journal of Business and Economic Statistics, 17(1), pp. 36–49.
100 SumHes
SumHes The Penn World Table, v. 5
Description

country : World
Usage
data(SumHes)
Format
year the year

country the country name (factor)
opec OPEC member?
com communist regime?
pop country’s population (in thousands)
gdp real GDP per capita (in 1985 US dollars)
sr saving rate (in percent)
Source
Online supplements to Hayashi (2000).

http://fhayashi.fc2web.com/datasets.htm
References
Hayashi, F. (2000) Econometrics, Princeton University Press, chapter 5, pp. 358–363.

Summers, R. and A. Heston (1991) “The Penn World Table (Mark 5): An expanded set of interna-
tional comparisons, 1950–1988”, Quarterly Journal of Economics, 106(2), pp. 327-368.
summary.plm 101
summary.plm Summary for plm objects
Description
The summary method for plm objects generates some more information about estimated plm mod-
els.
Usage
summary(object, vcov = NULL, ..., .vcov = NULL)
## S3 method for class 'summary.plm'
width = getOption("width"), subset = NULL, ...)
Arguments
object an object of class "plm",
x an object of class "summary.plm",
subset a character or numeric vector indicating a subset of the table of coefficients to
be printed for "print.summary.plm",
vcov a variance–covariance matrix furnished by the user or a function to calculate one
(see Examples),
.vcov deprecated, use argument vcov,
digits number of digits for printed output,
width the maximum length of the lines in the printed output,
Details
The summary method for plm objects (summary.plm) creates an object of class c("summary.plm", "plm", "panelmodel")
that extends the plm object it is run on with various information about the estimated model like (in-
ferential) statistics, see Value. It has an associated print method (print.summary.plm).
Value
An object of class c("summary.plm", "plm", "panelmodel"). Some of its elements are carried
over from the associated plm object and described there (plm). The following elements are new or
changed relative to the elements of a plm object:
fstatistic F statistic for joint test of significance of coefficients (class "htest") (robust F
statistic in case of supplied argument vcov, see pwaldtest for details),
102 summary.plm
coefficients a matrix with the estimated coefficients, standard errors, t–values, and p–values,
if argument vcov was set to non-NULL the standard errors (and t– and p–values)
in their respective robust variant,
vcov the "regular" variance–covariance matrix of the coefficients (class "matrix"),
rvcov only present if argument vcov was set to non-NULL: the furnished variance–
covariance matrix of the coefficients (class "matrix"),
r.squared a named numeric containing the R-squared ("rsq") and the adjusted R-squared
("adjrsq") of the model,
df an integer vector with 3 components, (p, n-p, p*), where p is the number of
estimated (non-aliased) coefficients of the model, n-p are the residual degrees
of freedom (n being number of observations), and p* is the total number of
coefficients (incl. any aliased ones).
Author(s)
Yves Croissant
See Also
plm for estimation of various models; vcovHC for an example of a robust estimation of variance–
covariance matrix; r.squared for the function to calculate R-squared; print.htest for some infor-
mation about class "htest"; fixef to compute the fixed effects for "within" (=fixed effects) models
and within_intercept for an "overall intercept" for such models; pwaldtest
Examples
summary(zz)
# summary with a funished vcov, passed as matrix, as function, and

# as function with additional argument
data = Grunfeld, model="within", effect = "individual")
summary(wi, vcov = vcovHC(wi))
summary(wi, vcov = vcovHC)
summary(wi, vcov = function(x) vcovHC(x, method = "white2"))
# extract F statistic
wi_summary <- summary(wi)
Fstat <- wi_summary[["fstatistic"]]
# extract estimates and p-values

est <- wi_summary[["coefficients"]][ , "Estimate"]
pval <- wi_summary[["coefficients"]][ , "Pr(>|t|)"]
# print summary only for coefficent "value"

print(wi_summary, subset = "value")
vcovBK 103
vcovBK Beck and Katz Robust Covariance Matrix Estimators
Description
Unconditional Robust covariance matrix estimators a la Beck and Katz for panel models (a.k.a.
Panel Corrected Standard Errors (PCSE)).
Usage
vcovBK(x, type = c("HC0", "HC1", "HC2", "HC3", "HC4"),
cluster = c("group", "time"),
diagonal = FALSE,
...)
Arguments
x an object of class "plm"
type one of "HC0", "HC1", "HC2", "HC3", "HC4",
cluster one of "group", "time"
diagonal a logical value specifying whether to force nondiagonal elements to zero
Details
vcovBK is a function for estimating a robust covariance matrix of parameters for a panel model
according to the Beck and Katz (1995) method, a.k.a. Panel Corrected Standard Errors (PCSE),
which uses an unconditional estimate of the error covariance across time periods (groups) inside
the standard formula for coefficient covariance. Observations may be clustered either by "group"
to account for timewise heteroskedasticity and serial correlation or by "time" to account for cross-
sectional heteroskedasticity and correlation. It must be borne in mind that the Beck and Katz for-
mula is based on N- (T-) asymptotics and will not be appropriate elsewhere.
The diagonal logical argument can be used, if set to TRUE, to force to zero all nondiagonal elements
in the estimated error covariances; this is appropriate if both serial and cross–sectional correlation
are assumed out, and yields a timewise- (groupwise-) heteroskedasticity–consistent estimator.
Weighting schemes are analogous to those in vcovHC in package sandwich and are justified the-
oretically (although in the context of the standard linear model) by MacKinnon and White (1985)
and Cribari-Neto (2004) (see Zeileis (2004)).
The main use of vcovBK is to be an argument to other functions, e.g. for Wald–type testing: ar-
gument vcov. to coeftest(), argument vcov to waldtest() and other methods in the lmtest
package; and argument vcov. to linearHypothesis() in the car package (see the examples). No-
tice that the vcov and vcov. arguments allow to supply a function (which is the safest) or a matrix
(see Zeileis (2004), 4.1-2 and examples below).
104 vcovBK
Value
An object of class "matrix" containing the estimate of the covariance matrix of coefficients.
Author(s)
Giovanni Millo
References
Beck, N. and Katz, J. (1995) What to do (and not to do) with time-series cross-section data in
comparative politics. American Political Science Review, 89(3), pp. 634–647.
Cribari-Neto, F. (2004) Asymptotic inference under heteroskedasticity of unknown form. Compu-
tational Statistics & Data Analysis 45(2), pp. 215–233.
Greene, W. H. (2003) Econometric Analysis, 5th ed., Prentice Hall/Pearson, Upper Saddle River,
New Jersey, Sec. 13.9.3, 13.9.7.
MacKinnon, J. G. and White, H. (1985) Some heteroskedasticity-consistent covariance matrix esti-
mators with improved finite sample properties. Journal of Econometrics 29(3), pp. 305–325.
Zeileis, A. (2004) Econometric Computing with HC and HAC Covariance Matrix Estimators. Jour-
nal of Statistical Software, 11(10), pp. 1–17. URL http://www.jstatsoft.org/v11/i10/.
See Also
vcovHC from the sandwich package for weighting schemes (type argument).
Examples
library(lmtest)
library(car)
data("Produc", package="plm")
zz <- plm(log(gsp)~log(pcap)+log(pc)+log(emp)+unemp, data=Produc, model="random")
## standard coefficient significance test
coeftest(zz)
## robust significance test, cluster by group
## (robust vs. serial correlation), default arguments
coeftest(zz, vcov.=vcovBK)
## idem with parameters, pass vcov as a function argument
coeftest(zz, vcov.=function(x) vcovBK(x, type="HC1"))
## idem, cluster by time period
## (robust vs. cross-sectional correlation)
coeftest(zz, vcov.=function(x) vcovBK(x,
type="HC1", cluster="time"))
## idem with parameters, pass vcov as a matrix argument
coeftest(zz, vcov.=vcovBK(zz, type="HC1"))
## joint restriction test
waldtest(zz, update(zz, .~.-log(emp)-unemp), vcov=vcovBK)
## test of hyp.: 2*log(pc)=log(emp)
linearHypothesis(zz, "2*log(pc)=log(emp)", vcov.=vcovBK)
vcovDC 105
vcovDC Double-Clustering Robust Covariance Matrix Estimator
Description
High-level convenience wrapper for double-clustering robust covariance matrix estimators a la

Thompson (2011) and Cameron, Gelbach and Miller (2011) for panel models.
Usage

vcovDC(x, type = c("HC0", "sss", "HC1", "HC2", "HC3", "HC4"),
...)
Arguments
x an object of class "plm" or "pcce"

type one of "HC0", "sss", "HC1", "HC2", "HC3", "HC4",
Details
vcovDC is a function for estimating a robust covariance matrix of parameters for a panel model with
errors clustering along both dimensions. The function is a convenience wrapper simply summing a
group- and a time-clustered covariance matrix and subtracting a diagonal one a la White.
The main use of vcovDC is to be an argument to other functions, e.g. for Wald-type testing: argument
vcov. to coeftest(), argument vcov to waldtest() and other methods in the lmtest package;
and argument vcov. to linearHypothesis() in the car package (see the examples). Notice that
the vcov and vcov. arguments allow to supply a function (which is the safest) or a matrix (see
Zeileis (2004), 4.1-2 and examples below).
Value
Author(s)
Giovanni Millo
106 vcovG
References
Cameron, A.C., Gelbach, J.B., & Miller, D.L. (2011) Robust inference with multiway clustering,
Journal of Business and Economic Statistics 29(2), pp. 238–249.
Thompson, S.B. (2011) Simple formulas for standard errors that cluster by both firm and time,
Journal of Financial Economics 99(1), pp. 1–10.
See Also
Examples
library(lmtest)
library(car)
zz <- plm(log(gsp)~log(pcap)+log(pc)+log(emp)+unemp, data=Produc, model="pooling")
coeftest(zz)
## DC robust significance test, default
coeftest(zz, vcov.=vcovDC)
coeftest(zz, vcov.=function(x) vcovDC(x, type="HC1", maxlag=4))
waldtest(zz, update(zz, .~.-log(emp)-unemp), vcov=vcovDC)
linearHypothesis(zz, "2*log(pc)=log(emp)", vcov.=vcovDC)
vcovG Generic Lego building block for Robust Covariance Matrix Estimators
Description
Generic Lego building block for robust covariance matrix estimators of the vcovXX kind for panel
models.
Usage
vcovG(x, type = c("HC0", "sss", "HC1", "HC2", "HC3", "HC4"),
cluster=c("group", "time"),
l=0,
vcovG 107
inner=c("cluster", "white","diagavg"),
...)
Arguments
cluster one of "group", "time",
l lagging order, defaulting to zero
inner the function to be applied to the residuals inside the sandwich: one of "cluster"
or "white" or "diagavg",
Details
vcovG is the generic building block for use by higher–level wrappers vcovHC, vcovSCC, vcovDC,
and vcovNW. The main use of vcovG is to be used internally by the former, but it is made available
in the user space for use in non–standard combinations. For more documentation, see see wrapper
functions mentioned.
Value
Author(s)
Giovanni Millo
References
Millo, G. (2014) Robust standard error estimators for panel models: a unifying approach Unpub-
lished manuscript.
See Also
vcovHC, vcovSCC, vcovDC, vcovNW, and vcovBK albeit the latter does not make use of vcovG.
Examples
zz <- plm(log(gsp)~log(pcap)+log(pc)+log(emp)+unemp, data=Produc,
model="pooling")
## reproduce Arellano's covariance matrix
vcovG(zz, cluster="group", inner="cluster", l=0)
## use in coefficient significance test
library(lmtest)
## define custom covariance function
## (in this example, same as vcovHC)
myvcov <- function(x) vcovG(x, cluster="group", inner="cluster", l=0)
108 vcovHC
## robust significance test

coeftest(zz, vcov.=myvcov)
vcovHC Robust Covariance Matrix Estimators
Description
Robust covariance matrix estimators a la White for panel models.
Usage
vcovHC(x, method = c("arellano", "white1", "white2"),
type = c("HC0", "sss", "HC1", "HC2", "HC3", "HC4"),
cluster = c("group", "time"), ...)
vcovHC(x, ...)
Arguments
x an object of class "plm" which should be the result of a random effects or a
within model or a model of class "pgmm" or an object of class "pcce",
method one of "arellano", "white1", "white2",
cluster one of "group", "time"
Details
vcovHC is a function for estimating a robust covariance matrix of parameters for a fixed effects or
random effects panel model according to the White method (White 1980, 1984; Arellano 1987).
Observations may be clustered by "group" ("time") to account for serial (cross-sectional) correla-
tion.
All types assume no intragroup (serial) correlation between errors and allow for heteroskedastic-
ity across groups (time periods). As for the error covariance matrix of every single group of ob-
servations, "white1" allows for general heteroskedasticity but no serial (cross–sectional) corre-
lation; "white2" is "white1" restricted to a common variance inside every group (time period)
(see Greene (2003, Sec. 13.7.1-2; 2012, Sec. 11.6.1-2) and Wooldridge (2002), Sec. 10.7.2);
"arellano" (see ibid. and the original ref. Arellano (1987)) allows a fully general structure w.r.t.
heteroskedasticity and serial (cross–sectional) correlation.
The main use of vcovHC is to be an argument to other functions, e.g. for Wald–type testing: ar-
vcovHC 109
A special procedure for pgmm objects, proposed by Windmeijer (2005), is also provided.
Value
An object of class "matrix" containing the estimate of the asymptotic covariance matrix of coeffi-
cients.
Author(s)
Giovanni Millo & Yves Croissant
References
Arellano, M. (1987) Computing robust standard errors for within-group estimators, Oxford Bulletin
of Economics and Statistics, 49(4), pp. 431–434.
New Jersey.
New Jersey.
Windmeijer, F. (2005) A finite sample correction for the variance of linear efficient two–step GMM
estimators, Journal of Econometrics, 126(1), pp. 25–51.
White, H. (1980) Asymptotic Theory for Econometricians, Ch. 6, Academic Press, Orlando (FL).
White, H. (1984) A heteroskedasticity-consistent covariance matrix and a direct test for heteroskedas-
ticity. Econometrica 48(4), pp. 817–838.
Wooldridge, J. M. (2002) Econometric Analysis of Cross Section and Panel Data, MIT Press, Cam-
bridge (MA).
See Also
Examples
library(lmtest)
library(car)
data = Produc, model = "random")
110 vcovNW

coeftest(zz)
## robust significance test, cluster by group
## (robust vs. serial correlation)
coeftest(zz, vcov.=vcovHC)
coeftest(zz, vcov.=function(x) vcovHC(x, method="arellano", type="HC1"))
## idem, cluster by time period
## (robust vs. cross-sectional correlation)
coeftest(zz, vcov.=function(x) vcovHC(x, method="arellano",
type="HC1", cluster="group"))
## idem with parameters, pass vcov as a matrix argument
coeftest(zz, vcov.=vcovHC(zz, method="arellano", type="HC1"))
waldtest(zz, update(zz, .~.-log(emp)-unemp), vcov=vcovHC)
linearHypothesis(zz, "2*log(pc)=log(emp)", vcov.=vcovHC)
## Robust inference for GMM models

data("EmplUK", package="plm")
ar <- pgmm(dynformula(log(emp) ~ log(wage) + log(capital) + log(output),
list(2, 1, 2, 2)), data = EmplUK, effect = "twoways",
model = "twosteps", gmm.inst = ~ log(emp),
lag.gmm = list(c(2, 99)))
rv <- vcovHC(ar)
mtest(ar, order = 2, vcov = rv)
vcovNW Newey and West(1987) Robust Covariance Matrix Estimator
Description
Nonparametric robust covariance matrix estimators a la Newey and West for panel models with
serial correlation.
Usage
vcovNW(x, type = c("HC0", "sss", "HC1", "HC2", "HC3", "HC4"),
maxlag=NULL,
wj=function(j, maxlag) 1-j/(maxlag+1),
...)
Arguments
type one of "HC0", "sss", "HC1", "HC2", "HC3","HC4",
vcovNW 111
maxlag either NULL or a positive integer specifying the maximum lag order before trun-
cation
wj weighting function to be applied to lagged terms,
Details
vcovNW is a function for estimating a robust covariance matrix of parameters for a panel model
according to the Newey and West (1987) method. The function works as a restriction of the Driscoll
and Kraay (1998) covariance to no cross–sectional correlation.
The main use of vcovNW is to be an argument to other functions, e.g. for Wald–type testing: ar-
Value
Author(s)
Giovanni Millo
References
Newey, W.K. & West, K.D. (1986) A simple, positive semi-definite, heteroskedasticity and auto-
correlation consistent covariance matrix. Econometrica 55(3), pp. 703–708.
See Also
Examples
library(lmtest)
library(car)
112 vcovSCC

coeftest(zz)
## NW robust significance test, default
coeftest(zz, vcov.=vcovNW)
coeftest(zz, vcov.=function(x) vcovNW(x, type="HC1", maxlag=4))
waldtest(zz, update(zz, .~.-log(emp)-unemp), vcov=vcovNW)
linearHypothesis(zz, "2*log(pc)=log(emp)", vcov.=vcovNW)
vcovSCC Driscoll and Kraay (1998) Robust Covariance Matrix Estimator
Description
Nonparametric robust covariance matrix estimators a la Driscoll and Kraay for panel models with
cross-sectional and serial correlation.
Usage

vcovSCC(x, type = c("HC0", "sss", "HC1", "HC2", "HC3", "HC4"),
cluster="time",
maxlag=NULL,
inner=c("cluster","white","diagavg"),
wj=function(j, maxlag) 1-j/(maxlag+1),
...)
Arguments

cluster switch for vcovG; set at "time" here,
maxlag either NULL or a positive integer specifying the maximum lag order before trun-
cation
inner the function to be applied to the residuals inside the sandwich: "cluster" for
SCC, "white" for Newey-West, ("diagavg" for compatibility reasons)
wj weighting function to be applied to lagged terms,
vcovSCC 113
Details
vcovSCC is a function for estimating a robust covariance matrix of parameters for a panel model
according to the Driscoll and Kraay (1998) method, which is consistent with cross–sectional and
serial correlation in a T-asymptotic setting and irrespective of the N dimension. The use with
random effects models is undocumented.
The main use of vcovSCC is to be an argument to other functions, e.g. for Wald–type testing:
argument vcov. to coeftest(), argument vcov to waldtest() and other methods in the lmtest
package; and argument vcov. to linearHypothesis() in the car package (see the examples).
Notice that the vcov and vcov. arguments allow to supply a function (which is the safest) or a
matrix (see Zeileis (2004), 4.1-2 and examples below).
Value
Author(s)
Giovanni Millo, partially ported from Daniel Hoechle’s (2007) Stata code
References
Driscoll, J.C. and Kraay, A.C. (1998) Consistent Covariance Matrix Estimation with Spatially De-
pendent Panel Data. Review of Economics and Statistics 80(4), pp. 549–560.
Hoechle, D. (2007) Robust standard errors for panel regressions with cross-sectional dependence.
Stata Journal, 7(3), pp. 281–312.
See Also
Examples
library(lmtest)
library(car)
coeftest(zz)
## SCC robust significance test, default
114 Wages
coeftest(zz, vcov.=vcovSCC)
coeftest(zz, vcov.=function(x) vcovSCC(x, type="HC1", maxlag=4))
waldtest(zz, update(zz, .~.-log(emp)-unemp), vcov=vcovSCC)
linearHypothesis(zz, "2*log(pc)=log(emp)", vcov.=vcovSCC)
Wages Panel Data of Individual Wages
Description
A panel of 595 individuals from 1976 to 1982, taken from the Panel Study of Income Dynamics
(PSID).
The data are organized as a stacked time series/balanced panel, see Examples on how to convert to
a pdata.frame.
observation : individuals
Usage
data(Wages)
Format
A data frame containing:
exp years of full-time work experience.

wks weeks worked.
bluecol blue collar?
ind works in a manufacturing industry?
south resides in the south?
smsa resides in a standard metropolitan statistical area?
married married?
sex a factor with levels "male" and "female"
union individual’s wage set by a union contract?
ed years of education.
black is the individual black?
lwage logarithm of wage.
within_intercept 115
Source

resourceId=13452
References
Cornwell, C. and P. Rupert (1988) “Efficient estimation with panel data: an empirical comparison
of instrumental variables estimators”, Journal of Applied Econometrics, 3(2), pp. 149–155.
Examples
# data set 'Wages' is organized as a stacked time series/balanced panel
Wag <- pdata.frame(Wages, index=595)
within_intercept Overall Intercept for Within Models Along its Standard Error
Description
This function gives an overall intercept for within models and its accompanying standard error
Usage
within_intercept(object, ...)
within_intercept(object, vcov = NULL, ...)
Arguments
object object of class plm which must be a within model (fixed effects model),
vcov if not NULL (default), a function to calculate a user defined variance–covariance
matrix (function for robust vcov),
... further arguments (currently none).
116 within_intercept
Details
The (somewhat artificial) intercept for within models (fixed effects models) was made popular by
Stata of StataCorp (see Gould (2013)) and later also adopted by Gretl (see Cottrell/Lucchetti (2016),
p. 160-161 (example 18.1)), see for treatment in the literature, e.g. Greene (2012), Ch. 11.4.4, p.
364. It can be considered an overall intercept in the within model framework and is the weighted
mean of fixed effects (see Examples for the relationship).
within_intercept estimates a new model which is computationally more demanding than just
taking the weighted mean. However, with within_intercept one also gets the associated standard
error and it is possible to get an overall intercept for twoway fixed effect models.
Users can set argument vcov to a function to calculate a specific (robust) variance–covariance matrix
and get the respective (robust) standard error for the overall intercept, e.g. the function vcovHC, see
examples for usage. Note: The argument vcov must be a function, not a matrix, because the model
to calculate the overall intercept for the within model is different from the within model itself.
Value
A named numeric vector of length one: The overall intercept for the estimated within model along
attribute "se" which contains the standard error for the intercept.
Author(s)
Kevin Tappe
References
Cottrell, A./Lucchetti, R., Gretl User’s Guide, http://gretl.sourceforge.net/gretl-help/
gretl-guide.pdf.
Gould, W. (2013), “How can there be an intercept in the fixed-effects model estimated by xtreg, fe?”,
http://www.stata.com/support/faqs/statistics/intercept-in-fixed-effects-model/.
Greene, W. H. (2012), Econometric Analysis, 7th ed., Prentice Hall, Ch. 11.4.4.
See Also
fixef to extract the fixed effects of a within model.
Examples
# estimate within model (unbalanced data)

mod_fe <- plm(mv ~ age + crim, data = Hedonic, index = "townid")
overallint <- within_intercept(mod_fe)
attr(overallint, "se") # standard error
# overall intercept is the weighted mean of fixed effects in the one-way case
weighted.mean(fixef(mod_fe), as.numeric(table(index(mod_fe)[[1]])))
# relationship of type="dmean", "level" and within_intercept in the one-way case

within_intercept 117
gi <- plm(inv ~ value + capital, data = Grunfeld, model = "within")

fx_level <- fixef(gi, type = "level")
fx_dmean <- fixef(gi, type = "dmean")
overallint <- within_intercept(gi)
all.equal(overallint + fx_dmean, fx_level, check.attributes = FALSE) # TRUE
# overall intercept with robust standard error

within_intercept(gi, vcov = function(x) vcovHC(x, method="arellano", type="HC0"))
Index
∗Topic aneweytest cipstest, 6

aneweytest, 4 mtest, 37
∗Topic array pbgtest, 41
detect_lin_dep, 9 pbltest, 43
∗Topic attribute pbsytest, 44
index, 22 pcdtest, 47
is.pbalanced, 23 pdwtest, 54
is.pconsecutive, 24 pFtest, 56
make.pbalanced, 29 phtest, 64
make.pconsecutive, 32 plmtest, 71
nobs, 38 pooltest, 77
pdim, 52 purtest, 84
plm.data, 70 pwaldtest, 89
punbalancedness, 81 pwartest, 91
pvar, 86 pwfdtest, 92
within_intercept, 115 pwtest, 94
∗Topic classes r.squared, 96
dynformula, 12 sargan, 98
lag, 28 ∗Topic manip
model.frame.pFormula, 36 detect_lin_dep, 9
pdata.frame, 50 pmodel.response, 76
pFormula, 55 ∗Topic model
pseries, 79 plm-package, 3
∗Topic datasets ∗Topic package
Cigar, 5 plm-package, 3
Crime, 8 ∗Topic piest
EmplUK, 13 piest, 66
Gasoline, 18 ∗Topic regression
Grunfeld, 19 ercomp, 14
Hedonic, 20 fixef.plm, 16
LaborSupply, 27 pcce, 45
Males, 35 pggls, 57
Parity, 40 pgmm, 59
Produc, 78 pht, 62
RiceFarms, 97 plm, 67
Snmesp, 99 plm-package, 3
SumHes, 100 pmg, 74
Wages, 114 pvcm, 87
∗Topic htest summary.plm, 101
118
INDEX 119
vcovBK, 103 Gasoline, 18

vcovDC, 105 Grunfeld, 19, 20
vcovG, 106
vcovHC, 108 has.intercept (plm), 67
vcovNW, 110 Hedonic, 20
vcovSCC, 112
[.pdata.frame (pdata.frame), 50 index, 22
[[.pdata.frame (pdata.frame), 50 is.pbalanced, 23, 31, 53
$.pdata.frame (pdata.frame), 50 is.pconsecutive, 24, 24, 29, 31, 33, 51
alias, 11 LaborSupply, 27
alias (detect_lin_dep), 9 lag, 26, 28, 29, 31, 33, 81
alias.lm, 10 lag.zoo, 29
aneweytest, 4 lead, 81
as.data.frame, 50 lead (lag), 28
as.data.frame.pdata.frame lm, 5, 14, 57, 60, 63, 66, 67, 74
(pdata.frame), 50
make.pbalanced, 24, 26, 29, 33
as.Formula.pFormula (pFormula), 55
make.pconsecutive, 24, 26, 30, 31, 32
as.list.data.frame, 51
Males, 35
as.list.pdata.frame (pdata.frame), 50
model.frame, 11, 55, 77
as.matrix.pseries, 29
model.frame (model.frame.pFormula), 36
as.matrix.pseries (pseries), 79
model.frame.pFormula, 36
Between, 29 model.matrix, 11, 55, 77
Between (pseries), 79 model.matrix (model.frame.pFormula), 36
between, 29 mtest, 37, 61
between (pseries), 79
nobs, 38, 53, 83
Between.pseries (pseries), 79
between.pseries (pseries), 79 Parity, 40
bgtest, 41, 42 pbgtest, 41, 45, 54, 92, 95
pbltest, 42, 43, 45, 54, 92, 95
Cigar, 5
pbsytest, 42, 44, 44, 54, 92, 95
cipstest, 6
pcce, 45
Crime, 8
pcdtest, 47
data.frame, 24, 26, 31, 33 pdata.frame, 23–26, 30–33, 46, 50, 52, 53,
detect_lin_dep, 9, 69 58, 74, 82, 83, 86, 88
diff, 81 pdim, 24, 26, 31, 33, 39, 51, 52, 82, 83, 87
diff (lag), 28 pdwtest, 42, 44, 45, 54, 92, 95
dynformula, 12, 61 pFormula, 37, 55
pFtest, 56, 73
EmplUK, 13 pggls, 57
ercomp, 14 pgmm, 26, 38, 59, 98
Extract, 50 pht, 62
phtest, 64
fixef, 70, 102, 116 piest, 66
fixef (fixef.plm), 16 plm, 10, 15, 17, 22, 26, 67, 96, 101, 102
fixef.plm, 16 plm-package, 3
Formula, 36, 37, 55, 76 plm.data, 70
formula.dynformula (dynformula), 12 plmtest, 45, 57, 71, 95
120 INDEX
plot.plm (plm), 67 summary.pht (pht), 62

pmg, 74 summary.piest (piest), 66
pmodel.response, 37, 55, 76 summary.plm, 69, 70, 90, 96, 101
pooltest, 77 summary.pmg (pmg), 74
print.dynformula (dynformula), 12 summary.pseries, 29
print.ercomp (ercomp), 14 summary.pseries (pseries), 79
print.fixef (fixef.plm), 16 summary.purtest (purtest), 84
print.htest, 102 summary.pvcm (pvcm), 87
print.panelmodel (plm), 67
print.pdata.frame (pdata.frame), 50 vcovBK, 103, 107
print.pdim (pdim), 52 vcovDC, 105, 107
print.purtest (purtest), 84 vcovG, 106
print.pvar (pvar), 86 vcovHC, 90, 102–108, 108, 109, 111, 113, 116
print.summary.aneweytest (aneweytest), 4 vcovNW, 107, 110
print.summary.fixef (fixef.plm), 16 vcovSCC, 107, 112
print.summary.pcce (pcce), 45
print.summary.pggls (pggls), 57 Wages, 114
print.summary.pgmm (pgmm), 59 waldtest, 90
print.summary.pht (pht), 62 Within, 29
print.summary.piest (piest), 66 Within (pseries), 79
print.summary.plm (summary.plm), 101 within_intercept, 17, 102, 115
print.summary.pmg (pmg), 74
print.summary.pseries, 29
print.summary.pseries (pseries), 79
print.summary.purtest (purtest), 84
print.summary.pvcm (pvcm), 87
Produc, 78
pseries, 24, 26, 31, 33, 79
punbalancedness, 24, 31, 33, 53, 81
purtest, 7, 84
pvar, 24, 26, 31, 33, 51, 53, 86
pvcm, 87
pvcovHC (vcovHC), 108
pwaldtest, 89, 101, 102
pwartest, 42, 44, 54, 91, 95
pwfdtest, 42, 44, 54, 92, 92, 95
pwtest, 94
r.squared, 96, 102

RiceFarms, 97
sargan, 61, 98
Snmesp, 99
SumHes, 100
summary.aneweytest (aneweytest), 4
summary.fixef (fixef.plm), 16
summary.pcce (pcce), 45
summary.pggls (pggls), 57
summary.pgmm (pgmm), 59

PLM

Uploaded by

Copyright:

Available Formats

PLM

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PLM

Uploaded by

Copyright:

Available Formats

Package ‘plm’

plm-package plm package: linear models for panel data

plm panel data estimators using lm on transformed data,

pgmm generalized method of moments (GMM) estimation for panel data,

pvcm variable coefficients models for panel data,

# replicates some results from Baltagi (2013), table 3.1

wi <- plm(inv ~ value + capital,

swar <- plm(inv ~ value + capital,

amemiya <- plm(inv ~ value + capital,

walhus <- plm(inv ~ value + capital,

aneweytest Chamberlain estimator and test for fixed effects

Angrist and Newey’s version of the Chamberlain test

aneweytest(formula, data, subset, na.action, index = NULL, ...)

Cigar Cigarette Consumption

state state abbreviation

x an object of class "pseries",

An object of class "htest".

data("Produc", package = "plm")

Crime Crime in North Carolina

detect_lin_dep Functions to detect linear dependence

## S3 method for class 'plm'

# setup a pFormula and a model frame

# no linear dependence in the pooling model's model matrix

# linear dependence present in the FE transformed model matrix

# look at the data: after FE transformation fact1 == -1*fact2

### Example 2 ###

Grunfeld$CEOtenure <- c(replicate(10, seq(from=s<-sample(1:10, 1), to=s+19, by=1)))

# look at the data

pform <- pFormula(inv ~ value + capital + CEOtenure + CEOage)

# no linear dependent columns in original data/pooling model

# CEOtenure and CEOage are linear dependent after FE transformation

# look at the transformed data: after FE transformation CEOtenure == 1*CEOage

dynformula Dynamic Formula

EmplUK Employment and Wages in the United Kingdom

ercomp Estimation of the error components

Amemiya, T. (1971) The estimation of the variances in a variance–components model, International

fixef.plm Extract the Fixed Effects

# relationship of type = "dmean" and "level" and overall intercept

# extract time effects in a twoways effects model

# with supplied variance-covariance matrix as matrix, function,

fx_level_robust3 <- fixef(gi, vcov = function(x) vcovHC(x, method = "white2"))

# calc. fitted values of oneway within model:

Gasoline Gasoline Consumption

Grunfeld Grunfeld’s Investment Data

Hedonic Hedonic Prices of Census Tracts in the Boston Area

mv median value of owner–occupied homes

Online complements to Baltagi (2001):

index Extract the indexes of panel data

is.pbalanced Check if data are balanced

punbalancedness for two measures of unbalancedness, make.pbalanced to make data balanced;

# take balanced data and make it unbalanced

is.pconsecutive Check if time periods are consecutive

## S3 method for class 'pdata.frame'

TRUE if the i-th individual has consecutive time periods,

# delete 2nd row (2nd time period for first individual)

# delete year 1937 (3rd period) for _all_ individuals

# panelmodel interface (first, estimate some models)

# default method for arbitrary vectors or NULL