The following pages were corrected in the third printing, which occurred in May, 2005. Many of these corrections were prompted by Jan Kiviet, to whom we are very grateful. pg009 Before the second-last display on the page, there was a slight change in wording to avoid possible confusion between the number of elements in a vector and its Euclidean length. (This change was made in the second printing.) In the last line of the subsection before "Multivariate Distributions", "an estimated parameter" was changed to "a parameter estimator" for clarity and correctness. pg018 There was a slight change in the third and fourth lines to eliminate possible ambiguity. pg021 In footnote 4, "x>0" replaces "a>0", which was incorrect. pg029 There were some changes in the first paragraph to better explain the notation for partitioned matrices. pg068 There was a minor change in the statement of the second part of Theorem 2.1 to make it clear that the theorem applies only to least squares residuals. pg095 The sentence that contains (3.17) was changed to make it clear that the matrix defined in that equation is finite. pg107 The last paragraph of Section 3.5 was changed to make it clear that the Gauss-Markov Theorem requires the regressors to be exogenous. pg113 The last sentence of the subsection on overspecification that ends on this page has been shortened because it was seriously misleading. pg129 There was a comma instead of a period at the end of equation (4.10). pg131 There was a comma instead of a period at the end of the second line of equation (4.13). pg142 The sentences between (4.33) and (4.34) were changed to avoid possible confusion. pg149 The sentences before and after the unnumbered equation prior to (4.46) were changed to make it clear that both this equation and (4.46) apply only when the random variables are uncorrelated. pg151 There have been changes following equation (4.50) to correct a false statement about why s^2 is consistent. pg154 There was a minor change below (4.58) to make it clear that r = k_2 in this case. pg168 In the second last sentence of the last full paragraph, a noncentrality parameter has been added in the notation for the noncentral chi-squared distribution. pg174 There were some changes in the statement of Exercise 4.11 to avoid making needlessly strong assumptions. pg211 In Exercise 5.12, ".95 level" replaced ".05 level", which was incorrect. pg221 There was a comma instead of a period at the end of equation (6.21). pg236 There was a missing period at the end of the second line of equation (6.50). pg239 In the equation prior to (6.57), a plim was added to make the equality hold exactly. pg244-6 There was a minor change following (6.73) to make it clear that the one-step and NLS estimators are only equivalent under the null hypothesis. Similar changes were made prior to (6.74) and (6.78). pg266 In the second full paragraph, a missing "a" was added before "new set". pg268 A typographical error was fixed three lines below (7.25). pg277 The first sentence of the very last paragraph was changed to make it clear precisely how the denominator of (7.47) is being changed. pg296 There is a slight change immediately after equation (7.80) to make it clear that the number of restrictions may be less than or equal to the number of regressors. pg306 In exercise 7.4, two missing commas were added in the covariances. pg319 The word "nonlinear" has been added in the first full paragraph on this page to avoid any possible misunderstanding. The second last full paragraph has been changed to avoid notational confusion. The matrix J^* is no longer referred to. There was also a change in the fourth printing. pg325-7 There have been several important changes on these three pages. Equation (8.39) and the text immediately following it on page 325 have been changed to make it clear that the instruments need only be predetermined, not necessarily exogenous. The last paragraph on page 326 has been substantially changed to correct an error. The first full paragraph on page 327 has been substantially changed to correct an error. pg331 Just below the middle of the page, sigma-acute has been replaced by s-acute, to conform to the notation used previously. pg349 In Exercise 8.16, there has been a slight change in notation to conform to the notation on page 331. pg355-6 A sentence was added at the end of the paragraph that includes (9.10) to clarify the nature of the efficient GMM estimator. This caused the page break between pages 355 and 356 to change. pg357-8 The first sentence on page 358 has been changed because it was not quite correct. This change affected the page break. pg360-1 Several typographical errors in equations (9.26) and (9.27) were corrected. Previously, some bars were in the wrong places. A similar correction was made just below (9.28) on page 361. In the line below (9.26), it is now made clear that EGMM means fully efficient GMM. pg414 The redundant plim was removed from equation (10.39). A sentence was added earlier in that paragraph to point out that the contributions to the gradient are uncorrelated. pg416 In the third line of the page, "in finite samples" has been added. pg422 Equation (10.59) and the text immediately following it have been changed so that no asymptotic analysis is involved. pg424 There was a minor addition just before (10.64), and there was a minor notational change in this equation, although the previous version was also correct. pg435-6 Just before (10.87), "(feasible)" has been added before GLS. A sentence has been added following (10.88) to make it clear that exogeniety is being assumed, and the paragraph was broken right afterwards for readability. pg438-9 A missing factor of 1/n has been added inside the large parentheses in expressions (10.92), (10.96), and (10.97). This omission effectively omitted a constant term from both loglikelihood functions. pg483 The sentence immediately prior to (11.67) has been changed to make it clear that this equation only involves observations which are observed. pg496 Exercise 11.5 has been changed slightly so that it is now the same as the exercise that was answered in the Instructor's Manual. pg499 In Exercise 11.23, the wrong equation number was cited. pg505-6 On page 505, a hat has been removed in equation (12.12). On page 506, the last sentence of the first paragraph has been changed slightly. pg508 There were two corrections on this page. In the first full paragraph, the second sentence has been changed to make it clear that unbiasedness requires that the regressors be exogenous. In the second full paragraph, near the end, "(or Delta)" has been added to make it clear that Sigma only needs to be known up to a scalar factor. pg531 Two lines below (12.77), a "greater than" sign, which was incorrect, has been replaced by a "greater than or equal to" sign. pg552 In Exercise 12.17, the x_{tj} variables have been changed to z_{tj} to accord with the notation of (12.54). In addition, the subscripts of the coefficient of y_{t1} in the second equation have been reversed. It is not the same as the coefficient of y_{t2} in the first equation. pg554 There were two errors in equation (12.127). The subscripts on x_{t5} were not actually subscripts, and the coefficient on q_t was incorrect. These errors have been corrected. pg559 Reference four lines from the bottom should be to Exercise 13.3 instead of 12.3. pg564-5 Equations (13.24) and (13.25) have been rewritten in a more illuminating way. The original versions were perfectly correct, but they were unnecessarily complicated. On page 565, The first six lines have changed because an error in the very first line was fixed. What rho(1) estimates is now stated correctly. The discussion just before (13.29) has been changed to make the definition of partial autocorrelation coefficients clearer. There was a slight change following (13.31) to make it clear that the number of parameters being counted does not include the variance of the innovations. pg578 The first paragraph on this page has been changed to avoid seeming to say that one can perform a unit root test as a standard t or F test by adding a forward reference to Chapter 14. pg589 At the end of the first paragraph on this page, there was a slight change to make it clear that sigma_t^2 is a conditional variance, as it was defined in (13.74). pg595 In the equation following (13.87), the upper limit of the second summation has been changed from m to g. pg599 The last sentence in Exercise 13.2 was changed to make it clearer just what is required. pg630 There were several corrections in the paragraph that contains equation (14.55). The number of endogenous variables should have been g, not k. The definition of the ECM estimator of eta_2 in the last sentence has been corrected. pg654 A typo has been corrected in the sentence, very near the middle of the page, that ends with "seen to be asymptotically valid." pg707 A very small typographical error was fixed in the second Donald-Paarsch reference.