Econometric Methods with Applications in Business and Economics

ERRATA LIST JANUARY 2010 (applies to books printed after mid 2005)

Chapter 2

Page Now is Should be

97 (-5/4) The least squares estimators a and b The least squares estimators a and b
given in (2.6) and (2.8) are linear given in (2.6) and (2.8) are linear
expressions in y1, …, yn. Such expressions in y1, …, yn. This can be
estimators are … verified from (2.20), which shows
that b =  c i y i , and (2.22), which
implies that a = d y
i i , where the
coefficients ci and di do not depend
on y1, …, yn. Such estimators are …
107 (+11) … x21 = 10 is in the middle … … x21 = 10 is near the middle …

Chapter 3

Page Now is Should be

126 (+5) If in addition Assumption 7 is If in addition Assumptions 2 and 7

satisfied, then … are satisfied, then …
139 (+2) … Exhibit 3.7 show that X’1 e = 0 … Exhibit 3.7 show that X’1 e = 0
(Panel 3), … and X’2 e = 0 (Panel 3), …
159 (-6) … 0.47 (shown in Panel 2), … … 0.47 (shown in Panel 2 of Exhibit
3.13), …
168 (+7) F(1, 469) distribution is 3.84, so … F(1, 469) distribution is 3.86, so …
181 (+18) <<in 3.6>> 3.6 ( Section 3.1.6) 3.6 ( Sections 3.1.6, 3.4.1)
181 (+25) <<in 3.6, just before a.>> add text Assume that all considered models
contain a constant term.
182 (-16) <<in 3.11>> … y1 = X1 + 1 with least squares
… y1 = X1 +  with least squares … …

Chapter 4

Page Now is Should be

221 (-8) 2(log(d))2 1

2(log(d))2
2
224 (+1) … distribution and of the t(5) … distribution and, as an example, of
distribution … the t(5) distribution …
246-249 <<The standard errors of the non- <<If ML standard errors are used,
 linear model reported in Panels 2 and then this has also consequences for
4 of Exhibit 4.19 are obtained from the t-statistics and Probability values
the NLS formulas on p. 206, see in Panel 2 and Panel 4 of Exhibit
(4.12), instead of the ML formulas 4.19 on p. 248. For comparison, the
on p. 228 and p. 243, see (4.57). As ML results are reported in two tables
ML uses the normal distribution, at the end of these errata of Chapter
these two standard errors are 4.>>
asymptotically the same, but they
differ in finite samples. Here, the
sample size of 12 is very small.>>
247 This corresponds to non-linear least This corresponds to non-linear least
(+6/7) squares. squares estimation of the
coefficients, but the ML standard
errors are based on ( 4.35) and (4.57)
instead of the NLS standard errors
that are based on (4.12).
247-249 <<The results of the Wald test are t1 = -1.119 and t2 = -0.911, and W1 =
based on the NLS standard errors. 2.225 and W2 = 1.107; none of these
The results shown on the right are test outcomes is significant at the 5
obtained by using the ML standard per cent significance level
errors. The outcomes differ quite
substantially, which is due to the
small number of 12 observations.>>
256 (-11) <<reduce spacing>> … larger errors for …
… larger errors for …
263 (+15) ̂ 2 = 4.49 ̂ = 4.49
264 (+3) PMLGMM
= 0.30 GMM
PML = 0.29
264 (+4) GMM
PML = 0.003 GMM
PML = 0.004
264 <<In Exhibit 4.21, Panel 4, standard 0.339 & 0.067
errors>>
0.334 & 0.066

Exhibit 4.19, Panel 2 and Panel 4, p. 249

Panel 2: Dependent Variable: LOGQ1 (brand 1)

Method: Maximum Likelihood
Included observations: 12
Convergence achieved after 61 iterations
LOGQ1 = C(1) + (C(2)/C(3)) * (D1^C(3)-1)
Parameter Coefficient Std. Error z-Statistic Prob.
C(1) 5.807116 0.038795 149.6884 0.0000
C(2) 10.29759 3.546852 2.903304 0.0037
C(3) -13.42942 12.00512 -1.118641 0.2633
Log likelihood 12.52991
Panel 4: Dependent Variable: LOGQ2 (brand 2)
Method: Maximum Likelihood
Included observations: 12
Convergence achieved after 42 iterations
LOGQ2 = C(1) + (C(2)/C(3)) * (D2^C(3)-1)
Parameter Coefficient Std. Error z-Statistic Prob.
C(1) 4.377873 0.039047 112.1195 0.0000
C(2) 10.27539 2.912369 3.528189 0.0004
C(3) -8.574115 9.411835 -0.910993 0.3623
Log likelihood 11.64128

Chapter 5

Page Now is Should be

304 (+16) … that 1 – 2 . The latter … … that 2 = 1 . The latter …

329 (-6) <<at end of displayed formula>> xi’ b*)2.
xi b*)2.
330 (-12) Formulate the model regression Formulate the regression model …
model …
344 (-7/6) (obtained by a regression over the nj (where ej is the sub-vector of e
observations in this group). containing the nj residuals in this
group).
348 The Likelihood Ratio test (5.39) … P The Likelihood Ratio test (5.39) is
(+10/21) = 0.000, so that homoskedasticity is based on four sums of squared
again rejected. residuals (SSR), that is, for the full
sample (with n = 474 employees) and
also for the three job categories
separately (with sample sizes n1 =
363, n2 = 27, and n3 = 84). The full
sample has SSR = 17.864 so that
s 2ML = SSR / n = 0.0377. For the three
job categories, we get SSR1 = 12.765
with s12,ML = SSR1 / n1 = 0.0352, SSR2
= 0.456 with s 22,ML = SSR2 / n2 =
0.0169, and SSR3 = 4.643 with
s 32,ML = SSR3 / n3 = 0.0553. With these
values, the LR-test is computed as LR
= 474 log(0.0377) – 363 log(0.0352)
– 27 log(0.0169) – 84 log(0.0553) =
14.39. With the (asymptotic) 2(2)
distribution, the P-value is 0.001, so
that homoskedasticity is again
rejected.
360 (+15) <<both terms (X’X)-1 in displayed 1 
1

formula>>  X' X  (two times)

n 
393 (+9) … wi = g(ei) and estimate … … wi = g(ei) / ei and estimate …
411 (-13) Hausman text … Hausman test …
429 (-7/6) <<in 5.11>> … terms i are IID(0,  2 ). By b we
… terms i are homoskedastic and denote …
uncorrelated. By b we denote …

Chapter 6

Page Now is Should be

458 outcomes provide some evidence for outcomes do not indicate the
(+1/2) the presence of heteroskedasticity (P presence of any heteroskedasticity.
= 0.01).
458 <<In Panel 1 of Exhibit 6.4, the line <<Delete these two lines from Panel
with Heteroskedasticity LM test 1 of Exhibit 6.4>>
value (df = 1), 6.237 & 6.186, and
the line with corresponding P-value,
0.0125 & 0.0129>>

Chapter 7

Page Now is Should be

570 (- 1 1
10/8) … success equal to . The first … success equal to . The first
2 2
model is preferred if B is model is preferred if B is
1 m
significantly larger than , and the significantly larger than , and the
2 2
second model is better if if B is second model is better if if B is
1 m
significantly smaller than . One significantly smaller than . One
2 2
can test the hypothesis … can test the hypothesis …
592(-2/1) … over time, and the forecast … over time, the forecast variance is
variance is constant for all horizons. constant for all horizons if the model
contains no ARMA terms, and this
variance converges to a finite limit
for larger horizons if the model
contains ARMA terms.
593 (+2) … variance increases for larger … variance increases without bound
horizons. for larger horizons.
Econometric Methods with Applications in Business and Economics

ERRATA LIST APRIL 2005

(applies only for books printed until mid 2005; corrected from mid 2005 onwards)

Chapter 2

Page Now is Should be

76 (-9) yi is denoted by RENDMARK and xi xi is denoted by RENDMARK and yi

by RENDCYCO by RENDCYCO
82 (+8/9) first and second sample moments first and second (non-centred)
sample moments
84 (+1) between y and x between x and y
96 (-7) ei = y – a – bxi ei = yi – a – bxi
98 (+7+9) <<replace symbol d by g (5 times)>>
(+7) di yi gi yi
(+7) d1 ,…, dn g1 ,…, gn
(+9) wi = di – ci wi = gi – ci
(+9) di = ci + wi gi = ci + wi

Chapter 3

Page Now is Should be

118 (+14) half of the variance can be explained half of the variability (as measured
by the variance) can be explained
127 (-5) sample mean sample average
152 (+3) Sections 1.4.1, 1.4.2. Sections 1.2.3, 1.4.1, 1.4.2.
152 (-6) (p. 31-5) (p. 32 and 34-5)
156 (+20) (3.31)). For (3.31) and (3.32)). For

Chapter 4

Page Now is Should be

194 (-6) first term first expression in the last line

194 (-5) fourth term fourth expression
211 (-4/2) , so that … we get , and using the fact that zh = eh +
<<NOTE: Reshuffle formulas and Xh^h it follows that ^h+1 = (Xh'Xh)-
text parts, but do not change 1
Xh'zh = (Xh'Xh)-1Xh'(eh + Xh^h) and
formulas! The formulas in the book hence
are correct, whereas those shown
here in the right column suffer
 from Word inconveniences>>
215 (+19) null hypothesis 0 holds true null hypothesis 2 = 0 holds true
217 (+12) ei = yi – f(xi, 1, 2) ei = yi – f(xi, b1, b2)
222 (+16) great variety of models, and large variety of models, and
applications models of special applications for models of special
interest interest
268 (+15) <<add text in Exercise 4.4, directly For simplicity, assume that x*, x and
above part (a); do not use new y are all IID (identically and
paragraph for this text>> independently distributed).
269 (+23) <<add ‘prime’ in Exercise 4.9.d>>
 = (0, 1, 0)  = (0, 1, 0)'
271 (-8) <<add text in Exercise 4.14, It is assumed that the error terms 
directly above part (a); do not use are NID(0, 2).
new paragraph for this text; NOTE:
normal size italic ‘sigma’ and
capital roman NID (not italic!) >>

Chapter 5

Page Now is Should be

370 (+24) between the residuals. Panel 5 between the residuals (see p. 362 and
Exercise 5.11). Panel 5
410 (+3) var(d)  var(…) var(d) = var(…)
<<NOTE: the  in lines 2, 4 and 5
are correct and should not be
changed>>
410 (-15) <<reduce space between parameter
and restriction>>
parameter restriction parameter restriction
433 (-15) <<in 5.25 (b)>> Check that the data in the data file
Sort the data with increasing values are sorted with increasing values of
of xi. x i.

Chapter 6

Page Now is Should be

485 (+4) <<numerator>> P[i  t] P[ –xi’ / < i  t]

485 (+4) <<numerator>> F(t) F(t) – F( –xi’ / )
485 (-7) <<numerator>> F((t – xi’) / ) F((t – xi’) / ) – F( –xi’ / )
485 (-5) <<at end formula (6.26) add, after
some open space and directly before
the closing dot, within parentheses as
shown>>
 ( yi > 0 )
493 (+15) <<displayed formula>>
E[ηi] = … = …+ λi E[ηi] = …= … + φi
<<SO: replace λi by φi>>
493 (+17) <<formula>> ωi = …= … – λi ωi = …= … – φi
<<SO: replace λi by φi>>
493 (+19) <<displayed formula>>
yi = …= yi = …=
… + σφ(xi’ / ) / (xi’ / ) + σωi … + σφ(xi’ / ) + σωi
<<SO: delete the denominator term
Φ(xi’ / ) of the fraction and also
the fraction line itself, and keep the
numerator term, now displayed on
normal line level>>
527 (-14) <<Exercise 6.9 (d)>>
(6.34) that implies (6.35) that implies (see p. 494)

Chapter 7

Page Now is Should be

582 (+14) E[ yt | y1 ] = (t-1)α E[ yt | y1 ] = y1 + (t-1)α

713 (+13) <<left column>> all the solutions of φ(z) = 0 lie
all the roots φ(z) = 0 lie outside outside