If the standard model accurately describes consumer behavior, then students buying a textbook should consider the by likelihood that a new edition will come out while the student is trying to sell the book. Chevalier and Goolsbee find that this is the case: students are less likely to purchase a textbook when the probability of a new edition arriving before the end of the semester is at its peak. In periods in which a book will certainly not be revised, book sales are relatively insensitive to price.

According to economists' standard model of consumer behavior in durable goods markets, the rational consumer will be forward-looking, and possible future events will affect today's buying decisions. But recent research in behavioral economics suggests that consumers are myopic and that they have a difficult time evaluating future costs and benefits -- they also are very impatient. Put differently, they have high short-run discount rates.

In **Are Durable Goods Consumers Forward Looking? Evidence from College Textbooks **(NBER Working Paper No. 11421), NBER researchers **Judith Chevalier** and **Austan Goolsbee** compare these two views by examining consumer purchasing behavior in the college textbook market. If the standard model accurately describes consumer behavior, then students buying a textbook should consider the by likelihood that a new edition will come out while the student is trying to sell the book. Chevalier and Goolsbee find that this is the case: students are less likely to purchase a textbook when the probability of a new edition arriving before the end of the semester is at its peak. In periods in which a book will certainly not be revised, book sales are relatively insensitive to price: a one percent increase in price reduces sales by only 0.9 percent. In a period in which the probability of revision is 50 percent, a one percent increase in price reduces sales by 2.3 percent. These d ata suggest that students are as responsive to a 1 percent change in the sticker price of the book today as they are to a 1 percent change in the expected resale price at the end of the semester. Thus, the results suggest that students pay close attention to future resale value and have low discount rates, characteristics in accord with the model.

It is popularly believed that publishers introduce new editions so that they can make more money by eliminating competition from used books. However, given that the authors' results show that "students are definitely forward-looking when they buy their textbooks" and that their behavior is consistent with very low discount rates, students will be less willing to buy a book that they expect will be revised quickly. The authors use their results about student demand to calibrate the effects of changing publishers' revision-time policies. The authors conclude that, "publishers of both economics and biology introductory textbooks would lose revenues by speeding up their revision cycles."

Textbooks are a significant cost for college students; some estimates suggest that textbook costs average $900 per year for students. However, students can defray some of these costs by purchasing used textbooks and/or by selling back their textbooks at the end of each semester. Some industry estimates suggest that the majority of students attempt to sell back most of their books every semester.

Using semester-by-semester data from 1698 college bookstores for 1997 to 2001 in the fields of psychology, biology, and economics, Chevalier and Goolsbee ask whether consumer purchasing behavior is consistent with students fully assessing their sellback opportunities when buying a new book. Their dataset contains information on both the textbooks that professors assign to their students and college bookstore sales. (Despite the growth in on-line commerce, the vast majority of college textbook purchases over the studied time period take place in college bookstores.) The authors estimate the factors that influence whether a student buys the assigned textbook.

Most college professors will be disappointed to learn that a significant fraction of students do not buy required textbooks; industry estimates suggest that, on average, approximately 20 percent of students do not buy a required textbook. Chevalier and Goolsbee estimate that the fraction of non-buyers increases with the price of the assigned textbook, as expected, and differs by field and with other characteristics of the book and the students assigned to read it. Most importantly, Chevalier and Goolsbee estimate the extent to which student purchase behavior responds to a rational expected resale price of the books.

The expected resale price that a student can obtain for a book has two components. First, students have to consider the price that their college bookstore will pay for their books. Typically, college bookstores buy back used textbooks at 50 percent of the new book price, while selling used textbooks of 75 percent of the new book price. Second, students have to consider the probability that their college bookstore will accept the book for resale.

Publishers revise textbooks every few years. Chevalier and Goolsbee show that iIntroductory books have a shorter lifespan before revision on average than more advanced ones, and that economics books have a shorter lifespan than psychology or biology textbooks. They show that the probability that the new edition will be introduced is essentially zero early in the life of the book, peaks when a book has been available for three to four years (depending on the field and level of the book), then falls off after that. Once a new textbook edition is introduced, "almost no one in the college bookstore supply chain is willing to buy or sell a used book for an outdated edition beyond one transitional semester" and the "buyback price for students holding the obsolete book essentially falls to zero.

-- Linda Gorman