Show Summary Details

Page of

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

date: 20 February 2020

(p. 661) Subject Index

(p. 661) Subject Index

Page numbers followed by f or t indicate figures or tables, respectively. Numbers followed by n indicate endnotes.

additive individual effects and time effects, 125–126
ADF (augmented Dickey-Fuller statistic), 57
advertising, 575
affine equivariance, 444n3
aggregate analysis, 410–412
aggregation, 411
agricultural economics, 364
Ahn, Lee, and Schmidt model, 532–533
Akaike Information Criterion (AIC), 143
α-trimmed mean, 444n2
ALSPAC (Avon Longitudinal Study of Parents and Children) (“Children of the 90s”), 498t, 505, 507
alternative asymptotics, 122–124
alternative bias correction method, 124
alternative procedures, 84–85
alternatives, irrelevant, 553
Anderson-Hsiao estimator, 344–345, 345f
applications, 453–641
approximate factor model, 9
area birth cohorts, 495–497
Arellano and Bond IV matrix, 439
Arellano-Bond estimator, 460
AR(1) errors, 562
arithmetic weighted average, 518
ARMA process, 487n5
AR(1) panel models, 446n13
AR(1) specifications, 552, 561
ASF (average structural function), 311
AS (nonlinear) GMM estimators, 98–99, 105
Asian countries, 536, 537, 538–539
asymptotics
alternative, 122–124
joint, 127–129
asymptotic standard errors, 80
attrition, 173, 192
augmented Dickey-Fuller statistic (ADF), 57
augmented regression model, 113
autocorrelation, 392, 393, 394, 395
autoregressive models, 113
integer-valued autoregressive model of order 1 (INAR(1)), 244
linear models with spatial autoregression, 195
average marginal effect (average treatment effect), 145n9
average regression coefficient, long-run, 47
average structural function (ASF), 311
average treatment effect (ATE), 145n9, 260, 311
dynamic (DATE), 271, 276
local (LATE), 260
for non-treated (ATENT), 260
for treated (ATET), 260
Avon Longitudinal Study of Parents and Children (ALSPAC) (“Children of the 90s”), 498t, 505, 507
bad leverage points, 418, 419f
Battese and Coelli model, 526–527
Bayes estimator, 409–410
Bayesian approach, 405–407
Bayesian Information Criterion (BIC), 112, 143, 538–539
Bayesian Stochastic Frontier Model, 533–534
BCS (Birth Cohort Study) (UK), 498t, 501, 508, 510–511
Berkson model, 350 (p. 662)
best linear unbiased predictor (BLUP), 394
β
fully modified OLS (FM-OLS) estimator of, 48
Within-OLS estimator of, 48–49
between equation, 379–380
BHPS (British Household Panel Survey), 172, 173, 192
bias, 332, 332f
of coefficient estimators, 98–99
within estimator, 332, 332f
of first differences and longest differences, 332, 332f
of fixed effect estimator, 124
Nickell bias, 85, 115–118, 455
bias-corrected estimators, 121, 409, 462–465
bias correction, 385
alternative method, 124
direct, 189
bias reduction
for binary choice models, 189–190
definition of, 142
bias stability assumptions, 264
BIC (Bayesian Information Criterion), 112, 143, 538–539
BI (bounded influence) estimators, 428–429
binary choice models, 175–192
binomial AR(1) panel model, 446n13
birth cohorts, 494, 495–497, 497–502, 498t, 503–506
Birth Cohort Study (BCS) (UK), 498t, 501, 508, 510–511
biweight function (Tukey bisquare weight function), 422, 423t, 424f
block missing, 152, 153f
BLUP (best linear unbiased predictor), 394
body mass index (BMI), 501
bootstrap method, 66, 366
Born-in-Bradford cohort, 498t, 505–506
Bounded Inefficiency Model, 534
bounded influence (BI) estimators, 428–429
Box-Cox transformed dependent variables, 309
Boyd Orr Cohort, 498t, 504, 508–509
Bradford Royal Infirmary, 505–506
brand choice models, 550–553
brand loyalty, 549, 572, 575
brand loyalty variable (GLijt), 572, 575
brand switching, 576
breakdown points (BDPs), 420–425
British Household Panel Survey (BHPS), 172, 173, 192
Burnside, Ethel Margaret, 503
Carnegie United Kingdom Trust, 504
CCE (Common Correlated Effects) estimation, 17
advantages of, 20
extension to non-/semi-parametric panel data models with large n and large t, 298
CCE (Common Correlated Effects) estimators, 17–21
application to unbalanced panels, 38–40
dynamic, 24–28
CCE Mean Group (CCEMG) estimators, 18–19, 23, 27–28, 29t
CCE Pooled (CCEP) estimators, 18–19, 27–28, 29t
CCR (conditionally correlated random) effects model, 241, 245
CDLM test, 31, 33
CDP test, 32, 34, 35, 36t, 38, 38t
CD tests, 38–40
Census of Fatal Occupational Industries, 602n2
Chamberlain estimator, 190
“Children of the 90s” (Avon Longitudinal Study of Parents and Children, ALSPAC), 498t, 505, 507
China, 537
choice, 552. see also T (number of choice situations or time series observations)
binary, 175–192
discrete choice models, 171–201, 548–582
lagged, 552, 569
multinomial, 174
ordered, 174, 190–191
choice probability, 554, 556
Cholesky decomposition, 287, 288
CIA (conditional independence assumption), 261–264, 280n16, 280n18
weak dynamic (W-DCIA), 273–274
cluster correction, 175–176 (p. 663)
clustering, 176
Cobb-Douglas distance function, 522
coefficient estimators, 98–99. see also specific estimators
cohort data, 493–516
cohort studies, 493–494
area birth cohorts, 495–497
birth cohorts, 494, 495–497
econometric methods applied to, 506–511
limitations to, 496
cointegration, 46–75
combining forecasts, 534–535
common cointegrating vectors, 50
Common Correlated Effects (CCE) estimation, 17
advantages of, 20
extension to non-/semi-parametric panel data models with large n and large t, 298
Common Correlated Effects (CCE) estimators, 17–21
application to unbalanced panels, 38–40
CCE Mean Group (CCEMG) estimators, 18–19, 23, 27–28, 29t
CCE Pooled (CCEP) estimators, 18–19, 27–28, 29t
dynamic, 24–28
common culture, 630
common factor models, 8–14
dynamic model with smooth factors, 163–168, 166t
dynamic model with stochastic factors, 163–168, 167t
static model with smooth factors, 163–168, 164t
static model with stochastic factors, 163–168, 165t
common factors, 482–485
assumptions typically made regarding, 8–9
deterministic, smooth factors, 162, 163f
generation procedures, 162, 163f
non-smooth factors, 162, 163f
observed (dt), 41n8
semi-strong, 12, 13–14
smooth factors, 162, 163–168, 163f, 164t
stochastic, 163–168, 165t
stochastic, non-smooth factors, 162, 163f
strong, 10–11, 12–13, 14
unobserved, 21
common-trend assumptions, 264, 265
composite quantile regression (CQR) method, 306
computational issues, 553–557
concentrated likelihood approach, 138–141
conditional independence, 177, 178
conditional independence assumption (CIA), 261–264, 280n16, 280n18
weak dynamic (W-DCIA), 273–274
conditional log likelihood function, 139
conditionally correlated random (CCR) effects model, 241, 245
conditional mean regression models, 286–302
conditional quantile regression models, 303–310
consistent estimation, 335–341
AH method, 344–345, 345f
constant-correlated effects, 86–90, 93
constant correlation, 79
consumer demand, 548–582
dynamics of, 549, 576
price elasticity of, 576, 578n8
sources of dynamics in, 572–576
state dependence in, 575
consumer demand models, 576–577
static, 577
typical structure of, 550–553
consumer learning models, 568
consumer taste, 575
contest function, 534–535
contiguity matrix, 195
continuously updated fully modified (CUP-FM) estimators, 52
control function procedures, 622–623
Cook’s distance, 446n16
Cornwell, Schmidt, and Sickles (CSS) panel stochastic frontier model, 523–524, 527
correlated (“fixed”) effects, 326
correlated random effects, 187, 191–192, 595–596
correlated random effects (CRE) model, 571
count-dependent variables, 233–256 (p. 664)
count models
AR(1) panel model, 446n13
individual effects in, 237
static, 237–243
count panel data, 233–256
country effects
cross-sectional fixed (exporter and importer), 632, 635n9
random, 629–631
country-pair fixed effects, 632, 635n11
country pairs
cross-section of, 610–613
three-way panels, 613–614
covariances, 477
CQR (composite quantile regression) method, 306
CRE (correlated random effects) model, 571
criminology, 214
cross-correlations, 477
cross-sectional fixed (exporter and importer) country effects, 632, 635n9
cross-sectional gravity models, 614–617, 622–623
cross-sectional independence, 30, 47–50, 316–318
cross-sectional interdependence, 625–626
cross-sectionally correlated panels, 51–55
cross-section data, 614–617
count data, 234–236
endogenous variables with, 623–624
grouped, 209
repeated, 624–625
repeated observation over time of, 613–614
two-way, 617–619
cross-section dependence, 374–376, 473–486
error, 28–38
large models with, 3–45
modeling, 474–476
panel data models with, 297–302
strong, 6
testing for, 33
tests for, 34, 393, 476–478
types of, 5–8
weak, 6, 7–8
cross-section regression models, 446n16
cross-sections
of country pairs, 610–613
repeated, 266–268, 617, 619–620
CSS (Cornwell, Schmidt, and Sickles) panel stochastic frontier model, 523–524, 527
culture, common, 630
CUP-FM (continuously updated fully modified) estimators, 52, 53
data. see also Panel data
cohort, 493–516
cross-section, 209
missing, 150, 152, 159, 168, 585, 617–619
degrees of freedom correction, 188–189
δi: fixed effects for, 208
demand
consumer demand, 548–582
good, 364
Denmark, 494
dependent data, 395
dependent variables
Box-Cox transformed dependent variables, 309
lagged, 21
deterministic, smooth factors, 162, 163f
DGP, 59
Dickey-Fuller coefficient
augmented (ADF), 57
modified, 56–57
differenced gravity equations, 622
difference-in-difference (DiD) approach, 264–268, 276–277
DIF (first-differenced) GMM estimators, 86, 98, 99, 105
directed tests, 477
disaggregate analysis, 410–412
disclosure avoidance, 353–354
discrete choice models, 171–201
computational issues, 553–557
of consumer demand, 548–582
estimation of, 553–564
extension to serially correlated taste shocks, 561–563
extension to state dependence, 563–564
general overview of, 553–557
other models, 184–185
over two alternatives, 174 (p. 665)
spatial panels and, 195–196
typical structure of, 550–553
discrete outcome models, 172–175
disturbances, 628–629
DOLS (dynamic OLS), 48–49
double fixed effects, 208–209
double-indexed variables, 630–631
double index process, 5
dummy variables, 178
Durbin–Hausman test, 62–63
Durbin regressors, 366
dynamic average treatment effect (DATE), 271, 276
dynamic average treatment effect (DATE) on the treated (DATET), 271
dynamic CCE estimators, 24–28
dynamic models, 76–110, 160–161, 455–469
additional parameters, 211–212
alternative procedures, 84–85
of discrete choice, 192–193
existing results, 96–97
with factor error structure, 21–28
fixed effects, 245–247
incidental parameters, 111–148
initial conditions for, 85–93
in labor economics, 597–599
latent class models, 252–253
linear, 438–441
literature review, 77–85
logit models, 214
measurement error models, 343–347
Monte Carlo design, 93–96
new results, 97–105
panel count models, 243–247
pooled, 244
prototype model, 113
random coefficient models, 407–410
random effects, 245
second-order, 214
simulation results, 93–105, 100t, 101t, 102t, 103t, 104t
with smooth factors, 163–168, 166t
with spatial errors, 385–389
with spatial lag, 382–385
spatial models, 364, 368–374, 382–391, 387t
specifications for, 243–244
static representation of, 456–458, 459t
with stochastic factors, 163–168, 167t
treatment models, 270–276
unbalanced, with interactive effects, 160–161
dynamic OLS (DOLS), 48–49
dynamic panel conditional logit estimators, 211–220
four periods or more with no regressor, 212–215
four periods or more with yT conditioned on, 218–219
four periods with the same last two-period regressors, 215–217
three periods or more using an estimator for δi, 219–220
three periods or more without yT conditioned on, 217–218
three periods or more with regressors, 217–220
dynamic panels, 586
nonlinear, 138–143
order selection in, 143–144
PC estimators for, 23–24
dynamics
in demand, 576
in labor economics, 599–601
in panel logit models, 211
in trade flows, 622
EC3SLS estimation, 395
ECHP (European Community Household Panel Survey), 172
econometrics, 583–607
economic growth. see also Productivity growth
decomposition of, 521–534
world, 535–539
economics, 548–549
efficiency
predictors of, 526–527
technical, 537
efficiency change
identified by regression, 521–534
index number approaches to, 520–521
Efficiency IV estimator, 524
efficient instrumental variables, 524 (p. 666)
ELFE (Étude Longitudinale Française depuis d’Enfance) (French Longitudinal Study of Children), 494
EM (Expectation-Maximization) algorithm, 158, 530
for factor models with missing data, 159
LS-EM-PCA estimation, 158–160, 168
for spatial panel data, 394
empirical Bayes estimator, 409
empirical Monte Carlo studies, 264
endogeneity, 594–599
endogeneous regressors, 622–625
endogenous attrition, 173
endogenous human capital, 586
endogenous spatial weights matrices, 393
endogenous variables
with cross-section data, 623–624
double-indexed, 630–631
with repeated cross-section (panel) data, 624–625
triple-indexed, 630
England. see United Kingdom
EQS software, 341
equilibrium, hedonic, 586–587
equivariance
affine, 444n3
crucial properties of, 444n3
of estimators, 421
to monotone transformations, 309
regression, 444n3
scale, 421, 425, 444n3
error correction model (ECMs), 371
vector error correction model (VECM), 485–486
error cross-sectional dependence tests, 28–38
errors
asymptotic standard errors, 80
factor error structure, 14–21, 21–28
homoskedastic, 33
idiosyncratic, 8–9
measurement, 325–362
one-way components, 151
panel corrected standard errors (PCSE), 474, 478–480
spatial, 385–389
weak or semi-strong factors in, 13–14
estimates
combining, 534–535
non-structural versus structural, 633–634
pooling of, 380
estimation
consistent, 335–341, 344–345, 345f
iterative, 154
nonparametric, 287
optimal, 345–347
parametric, 395–396
estimators. see also specific estimators
alternative, 472–473
coefficient, 98–99
general principles of, 472–473
logit, 202–232
Étude Longitudinale Française depuis d’Enfance (ELFE) (French Longitudinal Study of Children), 494
Euler equations, 351–353
European Community Household Panel Survey (ECHP), 172
European Union (EU), 632–633
exchangeable covariance matrices, 304
exogenous regressors, 339–341
Expectation-Maximization (EM) algorithm, 158, 530
for factor models with missing data, 159
LS-EM-PCA estimation, 158–160, 168
for spatial panel data, 394
explosive case, 372–373
exponential feedback, 243–244
exporter and importer (cross-sectional fixed) country effects, 632, 635n9
exporter-continent-time effects, fixed, 635n11
exporter-time fixed effects, 632
external information, 357–359
factor error structure
dynamic panel data models with, 21–28
large panels with strictly exogenous regressors and, 14–21
factor loadings, 8, 10–11, 533
factor models, 534. see also Common factor models
factors, 533. see also Common factors
fatal injury effects, 586–587 (p. 667)
feedback, exponential, 243–244
finance economics, 374
finite mixture models (FMMs), 249–252
first difference, 118–120, 430–432, 594
bias of, 332, 332f
first-differenced (DIF) GMM estimators, 86, 98, 99, 105
first-order spatial auto-regressive models, 9, 626
Fisher effects, 62
Fisher’s test statistic, 58
fixed effects (FE), 41n8, 85, 122, 123, 243, 355
additive, 125–126
bias B of, 124
country effects, 629–631
country-pair, 632, 635n11
double, 208–209
elimination of, 330–332
exporter-continent-time effects, 635n11
exporter-time, 632
formulations and problems of interest, 113–114
in gravity models, 629–631
importer-continent-time effects, 635n11
importer-time, 632
individual effects, 115–124, 377–378
individual-specific effects, 114
instrumental variables quantile regression with fixed effects (IVQRFE) estimator, 308
interactive, 126–129
limit distribution of, 121
versus random effects, 469, 629–631
in static models, 185–191
in static spatial models with serially correlated disturbances, 377–378
time-specific effects, 114
fixed effects models, 241–243, 293–297, 329–330
advantages of, 364
dynamic models, 245–247
logit models, 185, 187–188
probit models, 185, 188, 190
quantile regression models, 306–309
standard, 402
two-way, 611–612
fixed specification, 413–415
FMMs (finite mixture models), 249–252
FM-OLS estimator, 49, 52
forecasting, 394, 534–535
foreign direct investment models, 622–623, 626
foreign direct investment potentials, 633
forward orthogonal difference (FOD) transformation, 120, 390
FPCA (functional principal component analysis), 150, 155–156, 157–158
French Longitudinal Study of Children (Étude Longitudinale Française depuis d’Enfance, ELFE), 494
FSMQL (fully standardized Mallow’s type quasilikelihood approach), 441
F-test, 470–471
fully nonseparable models, 310
fully standardized Mallow’s type quasilikelihood approach (FSMQL), 441
functional principal component analysis (FPCA), 150, 155–156, 157–158
future research directions, 539–540
Gateshead Millennium Study (GMS), 498t, 506
Gauss-Chebyshev quadrature, 554
Gauss-Hermite quadrature, 554
Gauss-Legendre quadrature, 554
G-computation algorithm, 270
Gemini (birth cohort study), 498t, 506
generalized estimating equation (GEE), 182, 239
generalized least squares (GLS), 404, 405
breakdown point, 432
SUR-GLS estimator, 473–474, 480–482
generalized linear longitudinal mixed model (GLLMM), 441, 442
generalized linear models (GLMs), 235
generalized M-estimators (GM-estimators), 428–429, 432–433
generalized method of moments (GMM), 76–77, 89, 105–106, 318, 364, 366, 456
bias and precision of, 98–99
breakdown point, 432 (p. 668)
first-differenced (DIF) estimators, 86, 98, 99, 105
GMM-D estimator, 353
GMM-LN estimator, 353
LEV estimator, 86
merits of, 390–391
nonlinear (AS) estimators, 98–99, 105
robust (RGMM), 432, 433–434, 438–439, 439–440
in SDPD models, 389–391
for spatial panel data models, 393–394
system (SYS) estimators, 86, 98–99, 105
when T is large, 458–460
generalized PCSE estimator, 478
geometic weighted average, 518
German Socioeconomic Panel (GSOEP), 172, 173, 216
GHK algorithm, 550, 557–560, 564
Gibbs samplers, 407
GLLMM (generalized linear longitudinal mixed model), 441, 442
GLMs (generalized linear models), 235
global VAR (GVAR), 468, 485–486, 486–487
GLS (generalized least squares), 404, 405
breakdown point, 432
SUR-GLS estimator, 473–474, 480–482
GM-estimators (generalized-M estimators), 428–429, 432–433
GMS (Gateshead Millennium Study), 498t, 506
good demand, 364
good leverage points, 418, 419f
Granger causality, 468
gravity equations, 634n2
differenced, 622
dynamic adjustment, 622
systems of, 621
gravity models
balanced case, 614
cross-sectional, 614–617, 622–623
empirical topics, 614–634
fixed effects in, 629–631
for foreign direct investment, 622–623
fundamentals of, 609–610
of international trade, 608–641
random effects in, 629–631
repeated cross-sectional (three-way panel), 617, 619–620
in terms of relative log-odds or tetradic terms, 627
theoretical background, 609–610
with time variation, 613
two-way fixed effects model, 611–612
two-way random effects model, 612–613
Greene, Kumbhakar, and Tsionas latent class models, 528–530
grouped cross-section data, 209
group mean estimator, 409
growth convergence, 364
growth theory, 415, 518–519
GSOEP (German Socioeconomic Panel), 172, 173, 216
GVAR (global VAR), 468, 485–486, 486–487
HAC (heteroskedasticity and autocorrelation consistent) estimation, 395
Hampel M-estimator, 422, 423t, 424f
Hannan-Quinn (HQ) procedure, 143
HAS (Hertfordshire Ageing Study), 503
Hausman and Taylor type models
endogenous variables with cross-section data in, 623–624
endogenous variables with repeated cross-section (panel) data in, 624–625
Hausman specification test, 380–381, 412–413, 445n6
Hausman-Taylor estimator, 434–438
Hausman test statistic, 243
HCS (Hertfordshire Cohort Study), 503
health economics, 493–516
Health Survey for England, 503
hedonic equilibrium, 586–587
Helmert transformation, 390
Hertfordshire Ageing Study (HAS), 503
Hertfordshire Cohort Study (HCS), 503
Hertfordshire Studies, 498t, 503
heterogeneity, 412, 566
individual, 175–192
intercept, 589–591
intercept and slope, 587–593
types of, 573
unobservable, 174
heterogeneity tests, 412–413, 470–472 (p. 669)
heterogeneous panels, 472–473
heteroskedasticity, 394, 396
in gravity models of international trade, 614–617
spatial HAC estimation, 395
heteroskedasticity and autocorrelation consistent (HAC) estimation, 395
hidden Markov models, 252–253
hierarchical Bayes estimator, 409, 410
hierarchical models, 191–192
HILDA (Household Income and Labor Dynamics in Australia), 172
homogeneity, 411
homoskedastic errors, 33
homoskedastic slopes, 33
Horenstein, A. R., 41
hourly wage rates, 586–587
interactive factor model for, 600–601
Mincer wage models for, 589
Household Income and Labor Dynamics in Australia (HILDA), 172
housing prices, 395
HQ (Hannan-Quinn) procedure, 143
Huber M-estimator, 422, 423t, 424f
human capital, endogenous, 586
hurdle models, 175, 247–249
identification
nonparametric, 261–263, 268–269
semiparametric, 264–266
idiosyncratic errors, 8–9
idiosyncratic shocks, serially correlated, 563
idiosyncratic taste shocks, 552, 566
IIA (independence of irrelevant alternatives), 553
importer-continent-time effects, fixed, 635n11
importer-time fixed effects, 632
incidental parameters, 111–148
in discrete choice cases, 188–189
estimation of ρ with, 115–134
in micro panel data models, 202
testing for unit roots with, 134–138
incidental trends, 130–134
inconsistency, 328
incremental Sargan tests, 97
independence
conditional, 177, 178
cross-sectional, 316–318
of irrelevant alternatives (IIA), 553
of random terms in utility funcitons, 178
index i, 46
index number approaches, 520–521
indicator functions, 563
individual effects, 151
in count models, 237
fixed effects, 115–124, 125–126, 377–378
random effects, 378–379
individual heterogeneity, 175–192
individual-specific effects, 114, 202
infeasible Bayes estimator, 409
inference, valid, 478
inference approach, 395
influence functions, 420, 421, 422, 424f
influential observations, 418–450
innovation
decomposition of, 521–534
index number approaches to, 520–521
measurement of, 537
technical change due to, 526–527
instrumental variables (IVs), 76, 118–120, 268–270, 326
Arellano and Bond IV matrix, 439
efficient, 524
estimates of VSL, 594, 594t
ideal IV matrix, 391
robust IV estimators, 432–441
standard, 622–623
instrumental variables quantile regression with fixed effects (IVQRFE) estimator, 308
instruments
approaches for limiting the number of, 81
many, 80–81
proliferation of, 460–462
weak, 82–84
integer-valued autoregressive models of order 1 (INAR(1)), 244
integrated likelihood approach, 141–143
interactive effects, 149–170
fixed effects, 126–129
intercept and slope heterogeneity models, 591–593 (p. 670)
intercept heterogeneity models, 589–591
international trade models, 608–641
inventory, current, 577n3
inventory effects, 567, 568
inverse Mill’s ratio, 600–601
inverse normal test statistic, 58
inverse probability weighting, 192
Ireland, 494
IRI, 548
IRLS (iterated reweighted least squares), 430–431
irrelevant alternatives, 553
iterated reweighted least squares (IRLS), 430–431
iterative estimation, 154
IVQRFE (instrumental variables quantile regression with fixed effects) estimator, 308
JBFK test, 34, 35, 37t
Jensen’s inequality, 633
Jn: transformation approach with, 383–384
joint asymptotics, 127–129
KLIPS (Korean panel data), 216
Kneip, Sickles, and Song model, 530–532
Korean panel data (KLIPS), 216
Krugman-type models, 610
Kullback-Leibler approach, 112, 144
Kumbhakar model, 525
labor economics
dynamic models of, 597–599
endogeneity in, 594–599
framework for, 586–587
hedonic equilibrium in, 586–587
heterogeneous intercepts and slopes in, 587–593
intercept and slope heterogeneity models of, 591–593
intercept heterogeneity models of, 589–591
panel econometrics, 583–607
sample composition dynamics, 599–601
state dependence in, 597–599
summary, 601–602
labor-market data, 593–594
labor policy, 601–602
lagged choice, 569
lagged choice variables, 552
lagged dependent variables, 21
lagged prices, 567, 569
lagged purchases, 551, 568, 577n3
Lagrange multiplier (LM) test, 30–31, 363
LMAdj test, 31–32, 34, 35, 36t
LMS test, 33, 34, 35, 36t
robust modified versions, 471–472
for serial correlation and spatial autocorrelation, 392
for spatial effects, 392–393
standardized, 393
for unbalanced panels, 39–40
large N, small T panel data sets, 453–454
large panel data models, 3–45
large panel data sets, 444
large panels, 14–21
latent class models, 183, 249–252, 561, 562
dynamic, 252–253
general expression for, 249
of Greene, Kumbhakar, and Tsionas, 528–530
Latin American countries, 536, 537, 538–539
least squares. see also Ordinary least squares (OLS) estimators
of aggregate money demand function, 410–411, 410t
FPCA via (LS-FPCA), 157–158, 168
generalized (GLS), 404, 405, 432, 473–474, 480–482
iterated reweighted (IRLS), 430–431
LS-EM-PCA estimation, 158–160, 168
nonlinear (NLS), 393–394
robust and efficient weighted (REWLS) estimator, 431–432, 445n11
2SLS estimation, 393–394
least squares functional principal component analysis (LS-FPCA), 157–158, 168
Least Trimmed Squares (LTS) estimator, 425
reweighted LTS (RLTS), 431–432
LEV GMM estimator, 86
Life Study (UK), 498t, 502, 512n1
likelihood, 205
concentrated, 138–141
maximum likelihood (ML) estimator, 85
modified estimation equations, 189
quasi-maximum likelihood estimator (QMLE), 21–23
limited-dependent variables, 355–356
linearly approximated models, 627–628
linear models
dynamic models, 438–441
generalized (GLMs), 235
with spatial autoregression, 195
static models, 403–407, 419–432
LISCOMP model, 356
Lisrel model, 341–342
LISREL software, 341–342
literature reviews
for dynamic panel data models, 77–85
empirical findings on TFP growth, 536–539
empirical Monte Carlo studies, 264
empirical work on state dependence, 572–576
for missing data problem, 150
for panel stochastic frontier models, 533–534
for PCLEs, 208–209
LM (Lagrange multiplier) test, 30–31, 363
LMAdj test, 31–32, 34, 35, 36t
LMS test, 33, 34, 35, 36t
robust modified versions, 471–472
for serial correlation and spatial autocorrelation, 392
for spatial effects, 392–393
standardized, 393
for unbalanced panels, 39–40
local average treatment effect (LATE), 260
lock-in effects, 262–263
logit estimators, 202–232
longest differences, 332, 332f
longitudinal data sets, unbalanced, 173
longitudinal survey data sets, 172
long-run average regression coefficient, 47
loss functions, 422
Lothian Birth Cohorts, 498t, 503–504
LR statistic, 476
LSDV estimator, 98
LS-EM-PCA estimation, 158–160, 168
LS-FPCA (least squares functional principal component analysis), 157–158, 168
LTS (Least Trimmed Squares) estimator, 425
reweighted LTS (RLTS), 431–432
macroeconomic data, 453–492
macro shocks, 533
MAD (mean absolute deviation), 421
majority voting, 534–535
Malahanobis distance (or Rao’s distance), 429
Mallows’s estimators, 428–429
Mallow’s type quasilikelihood (MQL) approach, 441
many instruments problem, 460–462
marginal analysis, 238
marginal effects, 145n9, 209–210
marketing, 573
marketing mix variables, 565–566
Marketing Science, 548
market mapping, 553, 556
Markov chain Monte Carlo (MCMC) methods, 240–241
masking effects, 420, 425
matching, 267–268
MatLab software, 171, 487n1, 487n14
matrices
Arellano and Bond IV matrix, 439
endogenous spatial weights matrices, 393
exchangeable covariance, 304
spatial weight or contiguity matrix, 195
wage equation in matrix form, 595
maximum likelihood estimator (ML estimator or MLE), 85, 122, 465–467
Gaussian, 114
issues with, 390
partial, 208
Poisson quasi-MLE, 234–235
quasi-maximum likelihood estimator (QMLE), 21–23, 115, 118, 234–235
MCD (Minimum Covariance Determinant) estimator, 445n8
MCS (Millennium Cohort Study) (UK), 494, 498t, 501–502, 511, 512n1
MDE (minimum distance estimator)
mean absolute deviation (MAD), 421 (p. 672)
mean group estimator, 409, 470
measurement error, 325–362
basic results, 327–335
classical, 347–348
dynamic models of, 343–347
identification of, 334–335
in labor-market data, 593–594
multiplicative, 350–354
neglecting, 327–328
nonclassical, 347–350
nonlinear models of, 354–356
structural equation model (SEM) of, 342–343
Medical Expenditure Panel Survey (MEPS) (US), 172
M-estimators, 420–425
commonly used, 422, 423t
generalized (GM), 432–433
Hampel, 422, 423t, 424f
Huber, 422, 423t, 424f
influence functions of, 422, 424f
Tukey, 422, 423t, 424f
two-stage generalized (2SGM), 432–433
weight functions of, 422, 424f
method of moments (MOM) estimation, 363, 377–381
micro panel data models, 202
migration flows, 626, 634n2
Millennium Cohort Study (MCS) (UK), 494, 498t, 501–502, 511, 512n1
Mill’s ratio, inverse, 600–601
Mincer models, 584, 589
Minimum Covariance Determinant (MCD) estimator, 445n8
minimum distance estimator (MDE)
minimum distance theory, standard, 356
missing
block, 152, 153f
random, 152, 153f
regular, 152, 153f
missing data, 152, 168, 585, 617–619
EM algorithm for factor models with, 159
literature review, 150
missing observations, 168
in panel data, 152–154
simulation patterns, 162
for spatial panel data sets, 393–394
types of patterns, 152, 153f
mixed models, 183
mixture of logits model (MIXL), 560
normal (N-MIXL), 560, 561, 563
mixture-of-normals model, 561
MM-estimation, 426–427
model averaging, 534–535
models
binary choice models, 175–192
common factor models, 8–14
of consumer demand, 548–582
of cross-section count data, 234–236
of cross-section dependence, 474–476
discrete choice models, 171–201,548–582
discrete outcome models, 172–175
factor models with missing data, 159
gravity models of international trade, 608–641
nonlinear regression models, 172–175
panel data models and methods, 3–450
of world economic growth, 535–539
modified estimation (likelihood) equations, 189
moment conditions testing, 93
MOM (method of moments) estimation, 363, 377–381
money demand
least squares estimation of, 410–411, 410t
random coefficient estimation of, 411, 411t
monotone transformations, 309
Monte Carlo simulations
of dynamic panel data models, 93–96
empirical, 264
of unbalanced panel data models with interactive effects, 161–168
Moran I test, 392
mortality risk reduction: value of (VMRR), 584 (p. 673)
Mplus software, 341, 356
MQL (Mallow’s type quasilikelihood) approach, 441
MRC National Survey of Health and Development (NSHD) (UK), 498t, 499
MS-estimators, 425–428
multinomial choice, 174
multinomial logit estimators, 202–232
multiplicative measurement error, 350–354
Mundlak-Chamberlain model, 467
Mundlak-type models, 619–620
Mx software, 341
National Bureau of Economic Research, 517
National Child Development Study (NCDS) (UK), 498t, 499–500, 507–511
National Children’s Study (NCS) (US), 494
National Health Service (UK), 499
National Longitudinal Survey (NLS) (US), 172, 173, 583
National Survey of Health and Development (NSHD) (UK), 494, 498t, 499
natural experiments, 500, 585
NCDS (National Child Development Study) (UK), 498t, 499–500, 507–511
negative binomial (NB) models, 235, 236
neoclassical model, 518–519
new growth theory, 518–519
Nickell bias, 85, 115–118, 455
NLOGIT software, 171, 180
NLS (National Longitudinal Survey) (US), 172, 173, 583
N-MIXL (normal mixture of logits model), 560, 561, 563
noise, 83
adding, 563
signal-to-noise ratio (SNR), 95–96
noncointegration
null of, 56–64, 65–66
sequential approach to testing for, 61–62
nonlinear dynamic panels, 138–143
nonlinear (AS) GMM estimators, 98–99, 105
nonlinear least squares (NLS) method, 393–394
nonlinear measurement error models, 354–356
nonlinear panel data models, 176, 441–442
nonlinear regression models, 172–175
nonparametric identification, 261–263, 268–269
nonparametric regression models, 285–324
nonparametric tests, 313–318
nonseparable models, 310–313
non-structural estimates, 633–634
non-structural models, 633–634
normalized outcomes, 627
normal mixture of logits model (N-MIXL), 560, 561, 563
Northern Ireland. see United Kingdom
Norway, 494
notation, 46, 259–261, 280n6, 281n20, 281n25, 454
NSHD (National Survey of Health and Development) (MRC) (UK), 498t, 499
nuisance parameters, 81–82
null of noncointegration tests, residual-based, 56–64, 65–66
observables: selection on, 261–264, 600
observations
bad leverage points, 418, 419f
of cross-section data over time, 613–614
good leverage points, 418, 419f
influential, 418–450
missing, 152–154, 168
outliers, 442–444
random sampling of units, 178
repeated, 613–614
vertical outliers, 418, 419f
observed common factors (dt), 41n8
Occam’s razor, 578n5
odds ratio, 209–210
OECD (Organization for Economic Cooperation and Development) countries, 536, 537, 538–539
OIR (overidentifying restrictions) tests, 97, 99–105
OLS (ordinary least squares) estimators, 364 (p. 674)
asymptotic properties, 54
breakdown point, 432
dynamic OLS (DOLS), 48–49
FM-OLS estimator, 49, 52
pooled, 114
reduced form, 327–328
Within-OLS estimator, 48–49
OpenMx software, 341
optimal estimation, 345–347
ordered choice, 190–191
ordered logit, panel conditional, 220–223
ordered multinomial choice, 174
order selection, 143–144
ordinary least squares (OLS) estimators, 364
asymptotic properties, 54
breakdown point, 432
dynamic OLS (DOLS), 48–49
FM-OLS estimator, 49, 52
pooled, 114
reduced form, 327–328
Within-OLS estimator, 48–49
Organization for Economic Cooperation and Development (OECD) countries, 536, 537, 538–539
outliers, 442–444
overidentifying restrictions (OIR) tests, 97, 99–105
pairwise differences, 430–432, 445n9
panel cointegration, 46–75
panel conditional logit estimator (PCLE), 203, 230
dynamic, 211–220
with more than enough waves, 223
static, 204–211
panel conditional multinomial logit (PCML), 224–230
panel conditional ordered logit estimators, 220–223
panel corrected standard error (PCSE), 474, 478–480
panel count models
dynamic, 243–247
static, 237–243
panel data
applications, 453–641
compared to repeated cross-sections, 266–268
count, 233–256
detection of influential observations and outliers in, 442–444
large N, small T sets, 453–454
large sets, 444
macroeconomic, 453–492
measurement error in, 325–362
missing observations in, 152–154
repeated, 624–625
treatment effects and, 257–284
value of, 266–268
panel data models and methods, 3–450
applied to cohorts, 506–511
approximate factor model, 9
augmented regression model, 113
autoregressive model, 113
binary choice models, 175–192
common factor models, 8–14
for count-dependent variables, 233–256
with country-pair, exporter-time, and importer-time fixed effects, 632
with cross-sectional dependence, 297–302
discrete choice models, 171–201, 548–582
fixed effects models, 293–297
with fixed pair and fixed country-time effects, 632–633
gravity models of international trade, 608–641
heterogeneous, 14–15
with interactive effects, 151–152
with lagged dependent variables and unobserved common factors, 21
large models with cross-sectional dependence, 3–45
linear dynamic models, 438–441
linear static models, 419–432
micro models, 202
nonlinear, 176, 441–442
nonparametric regression models, 285–324
panel ordered logit models, 220–223
parametric models, 235–236 (p. 675)
prototype model, 113
random coefficient models, 402–417
random effects models, 286–293
regression models, 48–49, 57
robust methods, 418–450
simplest models, 233
spatial models, 363–401
static models, 364, 365–368, 367t
stochastic frontier models, 520, 523–524, 525, 526–527, 527–528, 528–530, 533–534
unbalanced models with interactive effects, 149–170
panel fully aggregated estimator (PFAE), 119
panel logit estimators, 202–232
panel regressions, 47–55
panels
cross-sectionally correlated, 51–55
cross-sectionally independent, 47–50
with fixed T, 118–120
heterogeneous, 472–473
with large T, 120–124
nonlinear dynamic, 138–143
pseudo-balanced, 482
random coefficient models in, 402–417
spatial, 195–196
three-way, 613–614, 617
unbalanced, 38–40, 192
Panel Study of Income Dynamics (PSID), 357, 583, 588
panel VAR cointegration tests, 66–70
parameters. see also specific parameters
nuisance parameters, 81–82
pooled versus individual specific, 469–473
testing for heterogeneity of, 470–472
parameter vector β
fully modified OLS (FM-OLS) estimator of, 48
Within-OLS estimator of, 48–49
parametric estimates, 395–396
parametric models, 235–236
Paris, France, 395
Park, Sickles, and Simar (PSS) models, 527–528
partial effects, 181–182
partially separable models, 310
partial MLE, 208
PCA (principal component analysis), 15–16, 150, 151, 483–484
for dynamic panels, 23–24
functional (FPCA), 150, 155–156, 157–158
with joint asymptotics, 127–129
LS-EM-PCA, 158–160, 168
LS-FPCA, 157–158, 168
PCLE (panel conditional logit estimator), 203, 230
dynamic, 211–220
with more than enough waves, 223
static, 204–211
PCML (panel conditional multinomial logit), 224–230
PCSE (panel corrected standard error), 474, 478–480
penalized criterion, 189
penalized IVQRFE estimator, 308
penalized QRFE (PQRFE) estimator, 307
Penn World Tables, 174
Perinatal Mortality Survey (PMS) (UK), 499–500
PFAE (panel fully aggregated estimator), 119
Phillips–Perron coefficient, 58
placebo tests, 262–263
PMS (Perinatal Mortality Survey) (UK), 499–500
Poisson generalized estimating equations (GEE) estimator, 239
Poisson models
fixed effects versions, 233
marginal density in, 187
random effects (RE) models, 239, 240
standard generalization of, 235–236
Poisson pseudo-maximum-likelihood (PPML) estimator, 615, 617
Poisson quasi-MLE estimator, 234–235
polynomial models, 354–355
poolability tests, 54, 314–316, 470–471
pooled dynamic models, 244
pooled least square (OLS) estimator, 114
pooled models, 238–239, 243
pooled panel data quantile regression models, 304–306
pooled parameters, 469–473
pooling of estimates, 380 (p. 676)
population-averaged models, 182, 190, 238–239
potential outcomes, 259, 277–279
PPML (Poisson pseudo-maximum-likelihood) estimator, 615, 617
PQRFE (penalized quantile regression fixed effects) estimator, 307
precision, 98–99
preferentialism, 630
preferential trade agreements, 631–633
pre-program tests, 262
price, 567
inventory effects on, 567, 568
lagged price, 567–568, 569
reference price effects, 567, 568
relative price, 578n6
as signal of quality, 567, 568
price coefficients, 551
price elasticity, 576, 578n8
price promotion, 575
principal component analysis (PCA), 15–16, 150, 151, 483–484
for dynamic panels, 23–24
functional (FPCA), 150, 155–156, 157–158
with joint asymptotics, 127–129
LS-EM-PCA, 158–160, 168
LS-FPCA, 157–158, 168
probit models
alternatives to, 560–561
correlated random effects (CRE) model, 571
discrete choice models, 554
fixed effects models, 185, 188, 190
random effects models, 179, 183–184, 554, 562
problems of interest, 113–114
product attributes, 577n1
production function, 519
productivity change, 537
productivity growth, 518–521, 537
productivity measurement, 517–547
propensity score, 264
prototype model, 113
pseudo-balanced panels, 482
pseudo-likelihood approaches, 442, 446n15
pseudo-panels, 496
PSID (Panel Study of Income Dynamics), 357, 583, 588
p-spacings, 477
public economics, 364
purchase carry-over effect, 573
purchases, lagged, 551, 568, 577n3
purchasing power parity (PPP) relations, 69
pure space recursive models, 364
Qj (true quality), 573
QML estimation, 393
QSF (quantile structural function), 311, 313
QTE (quantile treatment effect), 311
quadrature, 554–555
quality
price as signal of, 567, 568
quantile crossing, 309–310
quantile regression (QR) models
composite quantile regression (CQR) method, 306
conditional quantile regression models, 303–310
fixed effects panel data quantile regression models, 306–309
instrumental variables quantile regression with fixed effects (IVQRFE) estimator, 308
penalized quantile regression fixed effects (PQRFE) estimator, 307
pooled panel data quantile regression models, 304–306
quantile structural function (QSF), 311, 313
quantile treatment effect (QTE), 311
quasi-differencing approach, 126
quasilikelihood approach, 441
quasi-maximum likelihood (QML) estimation, 364
quasi-maximum likelihood estimator (QMLE), 21–23
Poisson, 234–235
random coefficient models, 402–417, 411t
random coefficients, 533
random effects (RE), 230, 243, 577n4
correlated, 187, 191–192, 595–596
in count data models, 184
country effects, 629–631
versus fixed effects, 469, 629–631 (p. 677)
in gravity models, 629–631
individual effects, 378–379
specification with fixed T, 387–388
in static models, 179–185
in static spatial models with serially correlated disturbances, 378–379
random effects assumption, 566
random effects models, 183, 239–241, 243, 286–293, 333–334
alternatives, 182–183
definition of, 179
dynamic models, 245
estimators for, 50
fully specified logit model, 184
generalizations of, 240–241
logit models, 179, 184
probit models, 179, 183–184, 554, 562
standard, 402
two-way random effects model, 612–613
random missing, 152, 153f
random sampling, 178
random specification, 413–415
random terms, 178
random utility, 171–172
random utility models, 172–173, 552
rank ordered logit models, 174
Rao’s distance (Malahanobis distance), 429
Rasch/Chamberlain estimator, 191
real estate economics, 364, 374, 395
reduced form (RF) parameters, 221
reference price effects, 567, 568
regional markets, 364
regression
for cross-sectionally correlated panels, 51–55
for cross-sectionally independent panels, 47–50
efficiency change identification by, 521–534
heterogeneous coefficients, 57
long-run average coefficients, 47
modified Dickey-Fuller coefficient, 56–57
Phillips–Perron coefficient, 58
seemingly unrelated (SUR), 150, 395
regression equivariance, 444n3
regression models
composite quantile regression (CQR) method, 306
cross-section, 446n16
for dynamic OLS (DOLS), 48–49
fixed effects panel data quantile regression models, 306–309
global VAR (GVAR), 485–486
with heterogeneous coefficients, 57
nonlinear, 172–175
nonparametric, 285–324
pooled panel data quantile regression models, 304–306
vector auto-regressive (VAR) models, 467–468
regressors
endogeneous, 622–625
strictly exogenous, 14–21
regularization parameters, 307
regular missing, 152, 153f
rejection frequency, 99
relative efficiency, 421
repeated cross-sections
compared to panel data, 266–268
data with, 617
three-way, 617, 619–620
repeated observations, 613–614
research directions, 539–540
residual-based tests, 56–64, 65–66
restrictions
benefits of, 336–337
reweighted LTS (RLTS) estimator, 431–432
REWLS (robust and efficient weighted least squares) estimator, 431–432, 445n11
RGMM (robust generalized method of moments estimator)
for linear dynamic models, 438–439, 439–440
for linear static models, 432, 433–434
ρ
estimation of, with incidental parameters, 115–134
limit distribution, 121
(concentrated) profile likelihood of, 123
QMLE of, 115
RLTS (reweighted LTS) estimator, 431–432
RND1 model, 536 (p. 678)
robust (term), 444n1
robust and efficient weighted least squares (REWLS) estimator, 431–432, 445n11
robust dispersion, 425–428
robust estimators
for linear dynamic models, 438–441
for linear static models, 419–432
for nonlinear models, 441–442
robust generalized method of moments estimator (RGMM)
for linear dynamic models, 438–439, 439–440
for linear static models, 432, 433–434
robust Hausman-Taylor estimator, 434–438
robust IV estimators, 432–441
robust LTS (RLTS) estimator, 430–432
robust methods, 418–450
robust pseudo-likelihood (RPL) estimator, 442
RPL (robust pseudo-likelihood) estimator, 442
R software, 171
R’s sem package, 341
Rubin rules, 358
SAH (self-assessed health), 192
sample selection, 173
sampling
GHK algorithm for, 564
for linear static models, 403–405
sequential importance sampling, 550
SAR (spatial auto-regressive) disturbances, 363
Sargan-Bhargava statistic, modified, 64
Sargan tests, 97
SAR (spatial auto-regressive) models, 395–396
first-order, 9, 626
linear, 195
semiparametric, 396
SAS software, 171, 180, 341
scale equivariance, 421, 425, 444n3
scanner data, 548, 549, 565–566
scanner data panels, 565
Scottish Mental Surveys, 503, 504
second-order dynamic models, 214
seemingly unrelated regression (SUR), 395
SUR-GLS estimator, 480–482
self-assessed health (SAH), 192
semiparametric and parametric estimates, 395–396
semiparametric identification, 264–266
semi-strong factors, 12, 13–14
SEMs (simultaneous equations models), 395
SEMs (structural equation models), 326, 341–343
sequential importance sampling, 550
serial correlation
strongly serially correlated case, 131–134
tests for, 392
weakly serially correlated case, 130–131
serially correlated disturbances, 377–381
serially correlated idiosyncratic shocks, 563
serially correlated taste shocks, 561–563
serially uncorrelated disturbances, 376–377
S-estimators, 425–428, 426t, 445n6
sieve bootstrap method, 66
signal, 83
signal-to-noise ratio (SNR), 95–96
simulations, 555–556
dynamic panel data model, 93–105
unbalanced panel data model with interactive effects, 161–168
simultaneous equations models (SEMs), 395
single index models, 185
SIPP (Survey of Income and Program Participation) (US), 172
skewness problem, 534
slope, homoskedastic, 33
SMA process, 366
smooth factors
deterministic, 162, 163f
dynamic models with, 163–168, 166t
static model with, 163–168, 164t
SNR (signal-to-noise ratio), 95–96
software, 341
Solow model, 89–90
Solow Residual (SR), 538
space-time filters, 368
spatial autocorrelation tests, 392, 393
spatial auto-regressive (SAR) disturbances, 363 (p. 679)
spatial auto-regressive (SAR) models, 395–396
first-order, 9, 626
linear, 195
semiparametric, 396
spatial cointegration, 371–372, 383, 387t
spatial Durbin regressors, 366
spatial dynamic panel data (SDPD) models
applications, 364
bias correction, 385
cases, 370
categories of, 364
error correction model (ECM) representation, 371
estimation and inference, 382–391, 387t
explosive case, 372–373, 384–385
fixed effects specification with fixed T, 388–389
general specifications, 368–369
GMM estimation, 389–391
with individual and time effects, 390
parameter spaces, 370, 371
pure unit root case, 386–387
QML estimation, 393
random effects specification with fixed T, 387–388
reduced form, 369
situations, 370
spatial cointegration case, 383, 387t
specifications, 368–374
stable case, 382–383, 387t
with time dummies, 383–384
unit root case, 373–374, 387t
spatial econometrics, 394
spatial effects, 392–393
spatial errors, 385–389
spatial heteroskedasticity and autocorrelation consistent (spatial HAC) estimation, 395
spatial lag (SL), 363, 382–385
spatial moving average (SMA) structures, 363–364
spatial panel data models, 363–401
spatial panels, 195–196
Spatial Stochastic Frontier model, 534
spatial weight matrix, 195, 393
specification tests, 183–184
sphericity test, 477
SPSS Amos software module, 341
spurious state dependence, 569
SR (Solow Residual), 538
stability theory, 444n1
standard errors, asymptotic, 80
Stata software, 171, 180, 208
external XTCSD procedure, 487n11
external XTLSDVC procedure, 487n1
SEM module, 341
state dependence, 552, 563–564, 586
in demand, 575
empirical work on, 572–576
in labor economics, 597–599
spurious, 569
testing for, 565–571
static models
case example, 151–152
of consumer demand, 577
count models, 237–243
fixed effects in, 185–191
random effects in, 179–185
with smooth factors, 163–168, 164t
spatial models, 364, 365–368, 367t, 376–381
with stochastic factors, 163–168, 165t
treatment models, 259–270
static panel conditional logit estimators, 204–211
stationarity, 371
statistical life value (VSL), 584, 585, 586, 588, 588t, 592–593, 592t, 594, 594t, 599, 601–602, 603nn4–5
steady state behavior: deviations from, 91–92
stochastic factors
dynamic model with, 163–168, 167t
non-smooth factors, 162, 163f
static model with, 163–168, 165t
stochastic frontier models, 520
additional models, 533–534
Battese and Coelli model, 526–527
Cornwell, Schmidt, and Sickles (CSS) model, 523–524, 527 (p. 680)
of Greene, Kumbhakar, and Tsionas, 528–530
Kumbhakar model, 525
in literature, 533–534
Park, Sickles, and Simar (PSS) models, 527–528
store choice, 578n6
strong cross-sectional dependence, 6
strong factors, 14
definition of, 10
example, 10–11
theorem 2, 12–13
strongly serially correlated case, 131–134
structural equation models (SEMs), 326, 341–343
structural estimates, 633–634
structural form (RF) parameters, 221
structural functions
average structural function (ASF), 311
quantile structural function (QSF), 311, 313
structural learning models, 568
structural models, 634
structural parameters, 202
sub-sampling algorithms, 445n7
SUR (seemingly unrelated regression), 395
SUR-GLS estimator, 473–474, 480–482
Survey of Income and Program Participation (SIPP, US), 172
swamping effects, 420
system estimator, 459–460
system (SYS) GMM estimators, 86, 98–99, 105
systems of equations, 621
T (number of choice situations or time series observations), 173
balanced, 173
large N, small T panel data sets, 453–454
sample log-likelihood function for general T, 218
sufficiently large, 455
unbalanced, 173
taste shocks
idiosyncratic, 552, 566
serially correlated, 561–563
technical efficiency, 537
technical innovation change, 526–527, 537
tests and testing. see also specific tests
for constant-correlated effects, 93
for cross-section dependence, 28–38, 33, 34, 476–478
directed tests, 477
for error cross-sectional dependence, 28–38
with incidental parameters, 134–138
for moment conditions, 93
for noncointegration, 61–62
overidentifying restrictions (OIR) tests, 97, 99–105
for panel cointegration, 55–70
residual-based, 56–64, 65–66
for sphericity, 477
for state dependence, 565–571
for unit roots, 134–138
TFP (total factor productivity) change, 537
TFP (total factor productivity) growth, 537
decomposition of, 522–523, 526–527, 530, 537
empirical findings, 536–539
measurement of, 518, 520–521, 528–529, 533–534
world growth findings, 536–539
third moments, 339
three-way panels, 613–614, 617
time dummies, 178, 383–384
time effects, 125–126, 474–476
time invariant individual variables (TIVs), 186
time series models, 243–244
time-space dynamics, 364, 368
time-space recursive models, 364
time-space simultaneous models, 364
time-specific effects, 114
time-varying parameters, 210–211
time-varying regressors, 215
TIVs (time invariant individual variables), 186
total factor productivity (TFP) change, 537 (p. 681)
total factor productivity (TFP) growth, 537
decomposition of, 522–523, 526–527, 530, 537
empirical findings, 536–539
measurement of, 518, 520–521, 528–529, 536
world growth findings, 536–539
trade cost variables, 635n6
trade data
cross-section (two-way), 617–619
missing, 617–619
repeated cross-sectional (three-way), 619–620
trade flows
effects of preferential agreements on, 631–633
normalized outcomes, 627
outcomes beyond goods trade, 610
potential, 633
trade models, 609
with country-pair, exporter-time, and importer-time fixed effects, 632
cross-sectional (three-way), 617, 619–620
cross-sectional fixed (exporter and importer) country effects models, 632
cross-sectional gravity models, 622–623
cross-sectional of country pairs, 610–613
double-indexed, 611
empirical topics, 614–634
with fixed pair and fixed country-time effects, 632–633
gravity models, 608–641
Hausman and Taylor type models, 623–624, 624–625
interpretation of disturbances, 628–629
Krugman-type, 610
linearly approximated models, 627–628
with multiple sectors, 610
Mundlak-type, 619–620
nonlinear, 635n11
outcomes beyond goods trade, 610
repeated cross-sectional (three-way), 617, 619–620
with three-way panels of country pairs, 613–614
triple-indexed, 610–611, 629, 630–631
two-part, 618, 619
two-way fixed effects model, 611–612
two-way random effects model, 612–613
transformation
based on X-differences, 465
forward orthogonal difference (FOD), 120, 390
Helmert, 390
transportation research, 364
t-ratio, 50
treatment effects, 257–284
average (ATE), 145n9, 260, 311
average for non-treated (ATENT), 260
average treatment effect for treated (ATET), 260
dynamic average (DATE), 271, 276
dynamic average on the treated (DATET), 271
dynamic models of, 270–276
identification of, 273–275
local average (LATE), 260
nonparametric identification of, 261–263, 268–269
quantile (QTE), 311
semiparametric identification of, 264–266
static models of, 259–270
trends, incidental, 130–134
triple-indexed trade models, 610–611, 629, 630–631
triple-indexed variables, 630
“True” Fixed Effects Model, 534
true quality (Qj), 573
t-test statistic, 56–57, 63–64
Tukey bisquare weight function (biweight function), 422, 423t, 424f
Tukey M-estimator, 422, 423t
tuning parameters, 307
2SGM (two-stage generalized M-estimator), 432–433, 438
2SLS estimation, 393–394
two-part models, 247–249
two-stage generalized M-estimator (2SGM), 432–433, 438
two-way (cross-section) data, 617–619
two-way fixed effects model, 611–612
two-way random effects model, 612–613 (p. 682)
unbalanced longitudinal data sets, 173
unbalanced panel data models with interactive effects, 149–170
data generation and implementation, 161–162
dynamic case, 160–161, 163–168, 166t, 167t
main findings, 163–168, 164t, 165t, 166t, 167t
Monte Carlo simulations, 161–168
static case, 151–152, 163–168, 164t, 165t
unbalanced panels, 192
UNIDO (United Nations Industrial Development Organization), 535–539
United Kingdom: cohort studies, 494, 497–502, 498t, 503–506, 511, 512n1
United Nations Industrial Development Organization (UNIDO), 535–539
United States, 494
unit roots
dynamic models with, 373–374, 386–387, 387t
testing for, 134–138
universal product code (UPC), 548
unobservable heterogeneity, 174
unobservables, 264–268, 268–270, 600
unordered multinomial choice, 174
UPC (universal product code), 548
U.S. Office of Management and Budget, 602
utility, 172–173
utility functions, 178
validation studies, 357–359
valid inference, 478
value of mortality risk reduction (VMRR), 584
value of statistical life (VSL), 584, 585, 586, 601–602, 603nn4–5
instrumental variables estimates of, 594, 594t
linear cross-section and panel data estimates of, 588, 588t, 603n5
panel quantile estimates of, 592–593, 592t
short-run estimates of, 599
steady state estimates of, 599
variable intercept models, 402, 415
variance–covariance (VC) matrices, 395
variance ratio (VR), 95
vector auto-regressive (VAR) models, 467–468
global (GVAR), 485–486, 486–487
vector error correction model (VECM), 50, 485–486
vertical outliers, 418, 419f
video games, 174
VMRR (value of mortality risk reduction), 584
voting, majority, 534–535
VSL (value of statistical life), 584, 585, 586–587, 601–602, 603nn4–5
instrumental variables estimates of, 594, 594t
linear cross-section and panel data estimates of, 588, 588t
panel quantile estimates of, 592–593, 592t
short-run estimates of, 599
steady state estimates of, 599
wage equation, 595
wage rates
effects of fatal injuries on, 586–587
interactive factor model for, 600–601
Mincer wage models for, 589
Wald tests, 63–64, 97, 99, 192, 392
Wales. see United Kingdom
Watson, M. W., 41
waves, more than enough, 223
weak cross-sectional dependence, 6, 7–8
weak dynamic conditional independence assumption (W-DCIA), 273–274
weak factors
definition of, 10
in errors, 13–14
example, 10–11, 12
theorem 2, 12–13
weak instruments, 82–84
weakly serially correlated case, 130–131
WGM-estimator (Within GM-estimator), 429
Wiener process, 56
Wishart distribution, 407
Within GM-estimator (WGM-estimator), 429
within-group estimator, 458, 459t (p. 683)
Within MS-estimator (WMS-estimator), 427–428, 429
Within-OLS estimator, 48–49
World Bank, 517
world economic growth models, 535–539
World Productivity Database (WPD) (UNIDO), 535–539
World Saving Data Base, 429
World Trade Organization (WTO), 631
WPD (World Productivity Database) (UNIDO), 535–539
X-differences, 119, 465
Young index, 521–522
zero-inflated models, 248
zeros, 617–619