Econometric Analysis - 7th Edition

Econometric Analysis, 7th Edition

ERRATA and DISCUSSION
Last updated October 14, 2012

Page Number Date Posted Errata

PAGE NUMBER DATE

erratum ... may be repeated for more than one on the page

Throughout 4/24/11

In the first printing of the book, the new references (15 pages of them, over 200) for Edition 7 were all omitted. They will be included in the second printing, June 2011.
Click here to download the missing references. (Right click to download) (The author, New York University and Tuba Suzer-Gurtekin, University of Michigan)

2 9/22/11

Thirteen lines up from the end of the page, "plateform" should be "platform.". (Erdogan CEVHER, The Ministry of Science, Industry and Technology, TURKEY)

5 5/4/11

In Figure 1.1, the labels of 2008 and 2009 are interchanged. (Rene Wells, Luxembourg)

6 9/22/11

Line 23, "processes" should be "process." (Erdogan CEVHER, The Ministry of Science, Industry and Technology, TURKEY)

9 9/22/11

Line 3 of Section 1.6.1, "provided" should be "provide." (Erdogan CEVHER, The Ministry of Science, Industry and Technology, TURKEY)

13 7/15/11

The second sentence in footnote 1 is missing its ending period. (Ali Tesiran, Middlesex University)

47 10/14/12

In the second line in the derivation of d, P'y should be P'X'y. (Samuel Mann, University of Cambridge)

72 4/24/11

In Table 4.4, in the first two rows, it should be noted that what is labeled "Standard Deviation" is actually the estimated standard error for the estimator. In the remaining rows, the standard deviation is computed for the R repetitions of the estimation. (Hai Anh, The World Bank)

76 9/3/11

In the last equation on the page, the X after the inverse matrix should be X'. (Randy Campbell, Mississippi State University)

87 4/1/11

In the description of the grid search algorithm, in the second line, the term mu^(0)=exp(x0'b), the function should not be exponentiated. (Author, New York University)

101 4/8/12

Two lines below equation (4-59), "Exercise 6" should be "Exercise 5." (Sangho Lee, Georgia State University)

119 4/1/11

In the sentence at the bottom of the page, it should be noted that the sign on the t ratio is lost. The absolute value of the t ratio is the square root of the F statistic. (Author, New York University)

193 11/3/11

The z statistic in the middle of the page should be 9.56299 without the negative sign. Note that the value is computed using the internal digits inside the computer program, not using the rounded values in the text. (Randy Campbell, Mississippi State University)

214 8/29/11

In the equation for the kernel weighted regression estimator at the top of the page, the "1/k" should be 1/h both times. Since the two cancel in the fraction, the result is unchanged. (William Greene, New York University)

260,261,263 4/1/11

On 260 at the end of (9-8), 261 in the first equation on the page and 263, second to last equation, the e_i should not be bold. (Hai anh Dang, World Bank)

272,275 4/1/11

The reference to White (1980a) on page 272 and to White (1980b) on page 275 should both be to (1980). (Johannes Pfeifer, Bonn Graduate School of Economics)

302-303 6/14/11

In table 10.1, the left most set of row labels, beginning with MW, are incorrect. They should be, in order, SW, CN, WC, MT, NE, MA, SO, MW. Some numerical corrections needed: In column alpha, 1.7834 should be 1.7835; 6.3783 should be 6.3784; in column beta1, 0.2948 ahould be 0.2949; -0.02040 should be -0.02041; in column beta6, -0.008143 should be -0.008144. (Tom Doan, Estima and the author.)

320 5/26/12

In the first row of G^-1, the middle term, b, should be b₂. In the upper left element of P', the closing parenthesis should be after b₂, not after a₁. (Inmaculada Alvarez Ayuso, Department of Economics, Autonoma de Madrid)

323 6/4/12

In the 5th equation on the page, the expression that follows "the matrix of reduced form coefficients, is observable,..." inside the first (leftmost) square brackets, X'Y should be X'X. (Fritz Sierra Tantaya, Cusco, Peru)

353 6/29/11

In (11-4), in the first line of the equation, in the inner summations, w^_ig should not be bold. (Hai anh Dang, World Bank)

359 6/21/11

In Table 11.4, in the first column of figures, 5.645 should be 5.635 and 0.178 should be 0.177. In the second column, 99.850 should be 94.635, 94.355 should be 99.850, 99.822 should be 99.823. (Tom Doan, Estima, Inc.)

369 6/29/11

On Table 11.5, in the Estimates column for FEVD, the coefficient on Blk should be -0.27507, not -0.14438 (Hai anh Dang, World Bank)

374 7/25/11

Consistent with the expression two equations above (11-33) on page 373, the entire right hand side of the expression for S_i^-1/2 should be multiplied by 1/s_e. (Stephanie Schurer, Victoria University of Wellington)

382 6/30/11

In the line after (11-48), the "(11-44)" should be "(11-47)." (Fred Dzanku, University of Reading)

383 6/29/11

In (11-51), in the matrix in square brackets, the first x_2,2 should be x_2,1. In the last row, the third and fourth terms should be x_n,2 and x_n,3 (Hai anh Dang, World Bank)

388 7/25/12

The first two sentences of Section 11.6.3 should be deleted. (Yupeng Zhang, School of Finance and Statistics, East China Normal University)

447 10/31/11

In the formula for the Epanechnikov kernel, |z| < 5 should be |z| < 2.236. (The square root of 5.) (Florian Wendelspiess Chavez Juarez, University of Geneva)

576 7/15/11

In the last column of Table 14.9, the value 0.838361 should be 0.12301719, as given earlier on page 378. The 0.838361 is the value for theta, also defined and reported on page 378. It should also be noted that the theta reported in the center column of results for the MLE, 42.66926, is the value of 1/var[u] as defined on page 575. (Tom Doan, Estima.)

598 4/22/11

In Table 14.16, the title should be "Estimated Latent Class Geometric Regression Model for DocVis." In Table 14.17, the three standard deviations should be, moving downward, 5.68979, 7.47579, 1.63076, respectively. (Tom Doan, Estima Inc. and the author, New York University)

614 4/24/11

In equation (15-8), the second plus sign should be a minus sign. (Hai Anh, The World Bank)

646-649 6/29/11

Example 15.16. The results in the example are correct as they stand, however, they are based on using the log of Koop and Tobias's log wage variable as the dependent variable. The LHS variable should not have been logged a second time. The full set of results in the application will change if the original variable is used instead. (Tom Doan, Estima, Inc.)

685 12/7/11

In the second line of the first equation, in the first term in the parentheses after the minus sign, the x should be w. ( Hoe Kyung Lee, Graduate School of Finance, KAIST, Korea.)

692 8/9/12

In equation (17-22), a q_i should appear after the plus sign in the parentheses. (Steven Yen, University of Tennessee, Knoxville)

695 9/5/12

In Table 17.1, in the column for Coefficient for the Complementary log log model, the constant term, 10.631, should be 10.031. (Author)

706 3/30/12

In the equation for the LM statistic at the top of the page, the second G should be G*'. (The * and prime are reversed.) The third X should be X*. The * is missing. (Greg Upton, Louisiana State University)

712 3/28/12

Several values in Table 17.6 are incorrect, as follows:
In row Constant: -1.13089 should be -1.13102
In row Income, 6 values should be 0.00027, 0.00014, 18.67, 0.00025, 0.00013, 19.4
In row Current Address, 6 values should be 0.00095, 0.00025, 3.86, 0.00094, 0.00024, 3.92
In row Self Employed, -0.47650 should be -0.47649, 0.05851 should be 0.05852, -25.66 should be -25.65
In row Major Derogs, -0.64792 should be -0.64791, 0.02525 should 0.02526. (Hui Wang, Louisiana State University)

715 4/21/11

Above Table 17.7, where the text states "Var[epsilon_i]=exp(...), it should state that the variance is the square of the exponent, as shown at the top of page 714. The results in the table are consistent with this formulation.

The partial effects in Table 17.7 appear to be incorrect. Using the average partial effects, the results in column 4 of the table should be as follows:
--------------------------------------------------------------------- Partial Standard Variable Effect Error |t| 95% Confidence Interval --------------------------------------------------------------------- AGE -.00779 .00263 2.96 -.01295 -.00263 INCOME .01709 .01566 1.09 -.01361 .04779 EDUC .03663 .00825 4.44 .02046 .05281 * KIDS -.16153 .04401 3.67 -.24779 -.07526 --------------------------------------------------------------------- * Dummy variable. Effect computed as first difference.
The results in the 6th column would be as follows:
--------------------------------------------------------------------- Partial Standard Variable Effect Error |t| 95% Confidence Interval --------------------------------------------------------------------- AGE -.00858 .00254 3.37 -.01356 -.00359 INCOME .06884 .02356 2.92 .02266 .11501 EDUC .02961 .00859 3.45 .01276 .04645 * KIDS -.16055 .04341 3.70 -.24563 -.07546 ---------------------------------------------------------------------
(Jean-Michel Benkert, University of Basel, Tom Doan, Estima, and the author)

725 6/29/11

The random effects logit results in table 17.8 are incorrect. The correct results are as follows:
----------------------------------------------------------------------------- Logit Model for Panel Data Dependent variable DOCTOR Log likelihood function -16277.04474 Estimation based on N = 27326, K = 7 Variance decomposition: Rho = .415027 [Rho = Var[u] / {Var[u] + Var[e]} Var[e] = pi-squared/3 = 3.289868 Var[u] = Var[e]*Rho/(1-Rho) = 2.334093 --------+-------------------------------------------------------------------- | Standard Prob. 95% Confidence DOCTOR| Coefficient Error z |z|>Z* Interval --------+-------------------------------------------------------------------- |Characteristics in numerator of Prob[Y = 1] Constant| .06449 .16391 .39 .6940 -.25678 .38576 AGE| .03416*** .00225 15.16 .0000 .02975 .03858 HHNINC| .00163 .11295 .01 .9885 -.21975 .22301 HHKIDS| -.26129*** .04588 -5.69 .0000 -.35122 -.17136 EDUC| -.05784*** .01071 -5.40 .0000 -.07883 -.03686 MARRIED| .02713 .05327 .51 .6106 -.07728 .13154 Sigma(u)| 1.52773*** .03284 46.53 .0000 1.46337 1.59209 --------+-------------------------------------------------------------------- Note: ***, **, * ==> Significance at 1%, 5%, 10% level. -----------------------------------------------------------------------------
(Tom Doan, Estima, Inc.)

806 4/1/11

Five lines up from the end of the page, there is a stray link to YouTube, "http://www.youtube.com/...embedded." that should not be there (and the author has no clue how it got there). (David Kessler, University of North Carolina)

810 4/21/11

In Table 18.14, the label theta should be an alpha. The values -0.3838 and -0.1359 in the 4th column should both be positive. (Tom Doan, Estima, Inc.)

850 11/15/11

In equation (19-13), "log" should be "ln." (Daniel Buncic, Institute of Mathematics and Statistics, University of St. Gallen, CH.)

875 10/12/12

The result in the first two equations on page 875 is correct as stated, as is the prelude on page 874. Footnote 27 is also correct, but confusing. On page 874, by theorem 19.5, l_i = f(-g'w_i/s_u)/[1-F(-g'w_i/s_u)] = f(g'w_i/s_u)/F(g'w_i/s_u) using the symmetry of the normal distribution. This second expression is f(-a_i)/F(-a_i). Let q_i = -a_i so that l_i = f(q_i)/F(q_i). The derivative is dE/dx_ik = b_k + b_l dl_i/dq_i dq_i/dx_ik. The last term is g_k/s_u. dl_i/dq_i = -q_il_i - l_i². This equals -(l_i² + q_il_i) = -(l_i² - a_il_i). Combining terms, dE/dx_ik = b_k - b_l(g_k/s_u)(l_i² - a_il_i) which is the result at the top of page 875. Footnote 27 is also correct, but confusing - the root problem is the proliferating minus signs. There is a typo in footnote 27; s_M should be s_u. Note, as well, in (19-22), the two occurrences of F should not be bold. (Author and Carol Troy, Feng-Chia University, Taiwan)

879 5/12/11

In Table 19.7, In column 5, 1.684 should be 1.680, 0.0964 should be 0.0963, 0.0837 should be 0.0836. In column 6, 0.449 should not be in parentheses - the parentheses should be around 0.318 in column 7. In the paragraph below the table, 0.064 should be 0.060 and -0.293 should be -0.279. It should be noted that slightly different answers are obtained depending on whether one computes the partial effects at the means of the data or computes the partial effects at the individual observations, then averages the partial effects. The values given here are based on the latter. (William Greene, NYU and Randy Campbell, Mississippi State University.)

1009 3/26/12

In equation (A-129), the numerator of the partial derivative should be x'A'. (Finn Martensen, Universitat Konstanz, Germany)

1029 09/3/11

At the top of the page, the hazard function is identified as a conditional probability. Professor Ali Tesaran observes "The hazard function (or rate) h(t) shows a risk. If we multiply it with dt, then the hazard probability is h(t)dt. Although both risk and probability cannot be less than 0, the risk can be greater than 1. (The word hazard generally implies the existence of a risk, and comes from the Arabic word "al zhar," meaning a die (page 72 in Tasiran. A. C. (1995) "Fertility Dynamics, Spacing and Timing of Births in Sweden and the United States", Contributions to Economic Analysis, 229, North-Holland.)" (Ali Tasiran, Birkbeck College)

1079 11/23/11

In the sentence above the second result in Theorem D.19, the nsigma_bar should be sqr(n), not n. (Citing Feller's (1968) Introduction to Probability Theory and Its Applications.) (Olivier Torres, University of Lille)

1110 5/11/11

In the item for Table F4.2, The Longley Data, "15 Yearly Observations" should be "16 Yearly Observations." (Randy Campbell, Mississippi State University)

Page Number	Date Posted	Errata
PAGE NUMBER	DATE	erratum ... may be repeated for more than one on the page
Throughout	4/24/11	In the first printing of the book, the new references (15 pages of them, over 200) for Edition 7 were all omitted. They will be included in the second printing, June 2011. Click here to download the missing references. (Right click to download) (The author, New York University and Tuba Suzer-Gurtekin, University of Michigan)
2	9/22/11	Thirteen lines up from the end of the page, "plateform" should be "platform.". (Erdogan CEVHER, The Ministry of Science, Industry and Technology, TURKEY)
5	5/4/11	In Figure 1.1, the labels of 2008 and 2009 are interchanged. (Rene Wells, Luxembourg)
6	9/22/11	Line 23, "processes" should be "process." (Erdogan CEVHER, The Ministry of Science, Industry and Technology, TURKEY)
9	9/22/11	Line 3 of Section 1.6.1, "provided" should be "provide." (Erdogan CEVHER, The Ministry of Science, Industry and Technology, TURKEY)
13	7/15/11	The second sentence in footnote 1 is missing its ending period. (Ali Tesiran, Middlesex University)
47	10/14/12	In the second line in the derivation of d, P'y should be P'X'y. (Samuel Mann, University of Cambridge)
72	4/24/11	In Table 4.4, in the first two rows, it should be noted that what is labeled "Standard Deviation" is actually the estimated standard error for the estimator. In the remaining rows, the standard deviation is computed for the R repetitions of the estimation. (Hai Anh, The World Bank)
76	9/3/11	In the last equation on the page, the X after the inverse matrix should be X'. (Randy Campbell, Mississippi State University)
87	4/1/11	In the description of the grid search algorithm, in the second line, the term mu^(0)=exp(x0'b), the function should not be exponentiated. (Author, New York University)
101	4/8/12	Two lines below equation (4-59), "Exercise 6" should be "Exercise 5." (Sangho Lee, Georgia State University)
119	4/1/11	In the sentence at the bottom of the page, it should be noted that the sign on the t ratio is lost. The absolute value of the t ratio is the square root of the F statistic. (Author, New York University)
193	11/3/11	The z statistic in the middle of the page should be 9.56299 without the negative sign. Note that the value is computed using the internal digits inside the computer program, not using the rounded values in the text. (Randy Campbell, Mississippi State University)
214	8/29/11	In the equation for the kernel weighted regression estimator at the top of the page, the "1/k" should be 1/h both times. Since the two cancel in the fraction, the result is unchanged. (William Greene, New York University)
260,261,263	4/1/11	On 260 at the end of (9-8), 261 in the first equation on the page and 263, second to last equation, the e_i should not be bold. (Hai anh Dang, World Bank)
272,275	4/1/11	The reference to White (1980a) on page 272 and to White (1980b) on page 275 should both be to (1980). (Johannes Pfeifer, Bonn Graduate School of Economics)
302-303	6/14/11	In table 10.1, the left most set of row labels, beginning with MW, are incorrect. They should be, in order, SW, CN, WC, MT, NE, MA, SO, MW. Some numerical corrections needed: In column alpha, 1.7834 should be 1.7835; 6.3783 should be 6.3784; in column beta1, 0.2948 ahould be 0.2949; -0.02040 should be -0.02041; in column beta6, -0.008143 should be -0.008144. (Tom Doan, Estima and the author.)
320	5/26/12	In the first row of G^-1, the middle term, b, should be b₂. In the upper left element of P', the closing parenthesis should be after b₂, not after a₁. (Inmaculada Alvarez Ayuso, Department of Economics, Autonoma de Madrid)
323	6/4/12	In the 5th equation on the page, the expression that follows "the matrix of reduced form coefficients, is observable,..." inside the first (leftmost) square brackets, X'Y should be X'X. (Fritz Sierra Tantaya, Cusco, Peru)
353	6/29/11	In (11-4), in the first line of the equation, in the inner summations, w^_ig should not be bold. (Hai anh Dang, World Bank)
359	6/21/11	In Table 11.4, in the first column of figures, 5.645 should be 5.635 and 0.178 should be 0.177. In the second column, 99.850 should be 94.635, 94.355 should be 99.850, 99.822 should be 99.823. (Tom Doan, Estima, Inc.)
369	6/29/11	On Table 11.5, in the Estimates column for FEVD, the coefficient on Blk should be -0.27507, not -0.14438 (Hai anh Dang, World Bank)
374	7/25/11	Consistent with the expression two equations above (11-33) on page 373, the entire right hand side of the expression for S_i^-1/2 should be multiplied by 1/s_e. (Stephanie Schurer, Victoria University of Wellington)
382	6/30/11	In the line after (11-48), the "(11-44)" should be "(11-47)." (Fred Dzanku, University of Reading)
383	6/29/11	In (11-51), in the matrix in square brackets, the first x_2,2 should be x_2,1. In the last row, the third and fourth terms should be x_n,2 and x_n,3 (Hai anh Dang, World Bank)
388	7/25/12	The first two sentences of Section 11.6.3 should be deleted. (Yupeng Zhang, School of Finance and Statistics, East China Normal University)
447	10/31/11	In the formula for the Epanechnikov kernel, \|z\| < 5 should be \|z\| < 2.236. (The square root of 5.) (Florian Wendelspiess Chavez Juarez, University of Geneva)
576	7/15/11	In the last column of Table 14.9, the value 0.838361 should be 0.12301719, as given earlier on page 378. The 0.838361 is the value for theta, also defined and reported on page 378. It should also be noted that the theta reported in the center column of results for the MLE, 42.66926, is the value of 1/var[u] as defined on page 575. (Tom Doan, Estima.)
598	4/22/11	In Table 14.16, the title should be "Estimated Latent Class Geometric Regression Model for DocVis." In Table 14.17, the three standard deviations should be, moving downward, 5.68979, 7.47579, 1.63076, respectively. (Tom Doan, Estima Inc. and the author, New York University)
614	4/24/11	In equation (15-8), the second plus sign should be a minus sign. (Hai Anh, The World Bank)
646-649	6/29/11	Example 15.16. The results in the example are correct as they stand, however, they are based on using the log of Koop and Tobias's log wage variable as the dependent variable. The LHS variable should not have been logged a second time. The full set of results in the application will change if the original variable is used instead. (Tom Doan, Estima, Inc.)
685	12/7/11	In the second line of the first equation, in the first term in the parentheses after the minus sign, the x should be w. ( Hoe Kyung Lee, Graduate School of Finance, KAIST, Korea.)
692	8/9/12	In equation (17-22), a q_i should appear after the plus sign in the parentheses. (Steven Yen, University of Tennessee, Knoxville)
695	9/5/12	In Table 17.1, in the column for Coefficient for the Complementary log log model, the constant term, 10.631, should be 10.031. (Author)
706	3/30/12	In the equation for the LM statistic at the top of the page, the second G should be G'. (The and prime are reversed.) The third X should be X. The is missing. (Greg Upton, Louisiana State University)
712	3/28/12	Several values in Table 17.6 are incorrect, as follows: In row Constant: -1.13089 should be -1.13102 In row Income, 6 values should be 0.00027, 0.00014, 18.67, 0.00025, 0.00013, 19.4 In row Current Address, 6 values should be 0.00095, 0.00025, 3.86, 0.00094, 0.00024, 3.92 In row Self Employed, -0.47650 should be -0.47649, 0.05851 should be 0.05852, -25.66 should be -25.65 In row Major Derogs, -0.64792 should be -0.64791, 0.02525 should 0.02526. (Hui Wang, Louisiana State University)
715	4/21/11	Above Table 17.7, where the text states "Var[epsilon_i]=exp(...), it should state that the variance is the square of the exponent, as shown at the top of page 714. The results in the table are consistent with this formulation. The partial effects in Table 17.7 appear to be incorrect. Using the average partial effects, the results in column 4 of the table should be as follows: --------------------------------------------------------------------- Partial Standard Variable Effect Error \|t\| 95% Confidence Interval --------------------------------------------------------------------- AGE -.00779 .00263 2.96 -.01295 -.00263 INCOME .01709 .01566 1.09 -.01361 .04779 EDUC .03663 .00825 4.44 .02046 .05281 * KIDS -.16153 .04401 3.67 -.24779 -.07526 --------------------------------------------------------------------- * Dummy variable. Effect computed as first difference. The results in the 6th column would be as follows: --------------------------------------------------------------------- Partial Standard Variable Effect Error \|t\| 95% Confidence Interval --------------------------------------------------------------------- AGE -.00858 .00254 3.37 -.01356 -.00359 INCOME .06884 .02356 2.92 .02266 .11501 EDUC .02961 .00859 3.45 .01276 .04645 * KIDS -.16055 .04341 3.70 -.24563 -.07546 --------------------------------------------------------------------- (Jean-Michel Benkert, University of Basel, Tom Doan, Estima, and the author)
725	6/29/11	The random effects logit results in table 17.8 are incorrect. The correct results are as follows: ----------------------------------------------------------------------------- Logit Model for Panel Data Dependent variable DOCTOR Log likelihood function -16277.04474 Estimation based on N = 27326, K = 7 Variance decomposition: Rho = .415027 [Rho = Var[u] / {Var[u] + Var[e]} Var[e] = pi-squared/3 = 3.289868 Var[u] = Var[e]Rho/(1-Rho) = 2.334093 --------+-------------------------------------------------------------------- \| Standard Prob. 95% Confidence DOCTOR\| Coefficient Error z \|z\|>Z Interval --------+-------------------------------------------------------------------- \|Characteristics in numerator of Prob[Y = 1] Constant\| .06449 .16391 .39 .6940 -.25678 .38576 AGE\| .03416* .00225 15.16 .0000 .02975 .03858 HHNINC\| .00163 .11295 .01 .9885 -.21975 .22301 HHKIDS\| -.26129* .04588 -5.69 .0000 -.35122 -.17136 EDUC\| -.05784* .01071 -5.40 .0000 -.07883 -.03686 MARRIED\| .02713 .05327 .51 .6106 -.07728 .13154 Sigma(u)\| 1.52773* .03284 46.53 .0000 1.46337 1.59209 --------+-------------------------------------------------------------------- Note: *, , * ==> Significance at 1%, 5%, 10% level. ----------------------------------------------------------------------------- (Tom Doan, Estima, Inc.)
806	4/1/11	Five lines up from the end of the page, there is a stray link to YouTube, "http://www.youtube.com/...embedded." that should not be there (and the author has no clue how it got there). (David Kessler, University of North Carolina)
810	4/21/11	In Table 18.14, the label theta should be an alpha. The values -0.3838 and -0.1359 in the 4th column should both be positive. (Tom Doan, Estima, Inc.)
850	11/15/11	In equation (19-13), "log" should be "ln." (Daniel Buncic, Institute of Mathematics and Statistics, University of St. Gallen, CH.)
875	10/12/12	The result in the first two equations on page 875 is correct as stated, as is the prelude on page 874. Footnote 27 is also correct, but confusing. On page 874, by theorem 19.5, l_i = f(-g'w_i/s_u)/[1-F(-g'w_i/s_u)] = f(g'w_i/s_u)/F(g'w_i/s_u) using the symmetry of the normal distribution. This second expression is f(-a_i)/F(-a_i). Let q_i = -a_i so that l_i = f(q_i)/F(q_i). The derivative is dE/dx_ik = b_k + b_l dl_i/dq_i dq_i/dx_ik. The last term is g_k/s_u. dl_i/dq_i = -q_il_i - l_i². This equals -(l_i² + q_il_i) = -(l_i² - a_il_i). Combining terms, dE/dx_ik = b_k - b_l(g_k/s_u)(l_i² - a_il_i) which is the result at the top of page 875. Footnote 27 is also correct, but confusing - the root problem is the proliferating minus signs. There is a typo in footnote 27; s_M should be s_u. Note, as well, in (19-22), the two occurrences of F should not be bold. (Author and Carol Troy, Feng-Chia University, Taiwan)
879	5/12/11	In Table 19.7, In column 5, 1.684 should be 1.680, 0.0964 should be 0.0963, 0.0837 should be 0.0836. In column 6, 0.449 should not be in parentheses - the parentheses should be around 0.318 in column 7. In the paragraph below the table, 0.064 should be 0.060 and -0.293 should be -0.279. It should be noted that slightly different answers are obtained depending on whether one computes the partial effects at the means of the data or computes the partial effects at the individual observations, then averages the partial effects. The values given here are based on the latter. (William Greene, NYU and Randy Campbell, Mississippi State University.)
1009	3/26/12	In equation (A-129), the numerator of the partial derivative should be x'A'. (Finn Martensen, Universitat Konstanz, Germany)
1029	09/3/11	At the top of the page, the hazard function is identified as a conditional probability. Professor Ali Tesaran observes "The hazard function (or rate) h(t) shows a risk. If we multiply it with dt, then the hazard probability is h(t)dt. Although both risk and probability cannot be less than 0, the risk can be greater than 1. (The word hazard generally implies the existence of a risk, and comes from the Arabic word "al zhar," meaning a die (page 72 in Tasiran. A. C. (1995) "Fertility Dynamics, Spacing and Timing of Births in Sweden and the United States", Contributions to Economic Analysis, 229, North-Holland.)" (Ali Tasiran, Birkbeck College)
1079	11/23/11	In the sentence above the second result in Theorem D.19, the nsigma_bar should be sqr(n), not n. (Citing Feller's (1968) Introduction to Probability Theory and Its Applications.) (Olivier Torres, University of Lille)
1110	5/11/11	In the item for Table F4.2, The Longley Data, "15 Yearly Observations" should be "16 Yearly Observations." (Randy Campbell, Mississippi State University)