This book is probably the most praised of all in the pool of undergraduate econometrics texts. It is accessible to a fairly broad range of students with quite different majors (economics, business school, MBA, ...). I believe there is one crucial problem with Gujarati's text: in introductory chapters it does not explain the sampling process from the statistical point of view, which forms a basis for a more theoretical but still applied approach to studying econometrics. Consequently, the adopted path in the book is necessarily limited to a fairly cook-book approach. Which is attractive for business or MBA students but shouldn't really be praised by would-be economists.
I would prefer Modern Econometrics: An Introduction by Thomas or even Undergraduate Econometrics by Hill/Griffiths/Judge that essentially covers the same material but is not that fat.
This book, although much-used, presents considerable problems from the point of view of the student using this as an introductory text for econometrics.First, the layout is quite dense and does not provide much guidance as to the relative importance of results. It would be far more helpful for the presentation to reflect the logical buildup of an argument to its conclusion.
Instead, although the verbal exposition is usually quite clear, the presentation is a jumble, with many relatively lengthy derivations placed in footnotes! This is bad pedagogigally, as it encourages the student to skip over what are useful (and often not too difficult)parts.
Second, and worse, is the continued use of deviation notation (ie. expressing a random variable in terms of deviation from its expected value) throughout the book. For this reason alone, I would never recommend this textbook for any class (incidentally, can anyone think of another textbook that follows this bizarre notation)? This is because, for the student who wishes to work through all derivations (which I always encourage) it imposes extra time constraints and a need to flick back through the book to check how variables are defined. This is incredibly frustrating and demotivating, and puts many students off econometrics before they've even given it a fair try.
Third, the book reveals what is (to my mind) an unhealthy preoccupation with estimation issues, as opposed to those of data quality. As people like Granger have consistently pointed out, the real issues in 21st century econometrics have to do with what sort of data we have, and what methods are most appropriate in different situations. Gujurati, partly reflecting its long-standing use, merely dwells on the iid specification. However, I should point out in its defence that this third defect is shared by most other introductory (and graduate) textbooks on econometrics.
As stated above, I would never recommend this book, preferring instead Hill's "Introductory Econometrics" which both motivates the student and takes them through the steps and methods they will need to adopt in further econometric studies, and always provides a good reference when it omits proofs and other details.
