The act of trading in a used car as partial payment for a new car resonates with practically all consumers. Such transactions are prevalent in many other durable goods markets ranging from golf clubs to CT scanners. What roles do trade-ins play in these markets? What motivates the seller to set up a channel to facilitate trade-ins? Intuitively, accepting a trade-in would appear to stimulate demand for the producer's product, but facilitating the resale of these used goods that substitute for new goods might also increase cannibalization. Although such transactions involve billions of dollars, we know little about this practice from the extant research literature.This paper develops an analytical model that incorporates key features of real-world durable goods markets; a) co-existence of new and used goods markets, b) consumer heterogeneity with respect to quality sensitivity, c) firms who anticipate the cannibalization problem arising from the co-existence of new and used goods, and d) lemons problems in resale markets, whereby sellers of used goods are better informed than buyers about the quality of their particular item.In our analysis, a trade-in policy amounts to an intervention by the firm in the used good market, which reduces inefficiencies arising from the lemons problem. It motivates owners to purchase new goods and reduces their proclivity to hold on to purchased goods because of the low price the latter would fetch in a lemons market.We also show that trade-in programs are more valuable for less reliable products because of the more acute lemons problem. Such programs increase the average quality of used goods offered for resale, which in turn increases used goods prices as well as the volume of transactions in the resale market. Trade-in programs are also more valuable for products that deteriorate more slowly, because the near-new quality of a used good allows it to compete more effectively for new good purchases.We test the key predictions of the model about price and volume of trade by assembling a dataset of transactions of US automobile consumers, and find broad support for our model. In particular, we find that a consumer buying a automobile with a trade-in receives an average discount of $644 (net of the value of the traded vehicle), and that this discount is larger for more unreliable make-models ($1,217) as well as for make-models that deteriorate slower ($1,251). The volume of used good trade is larger for more reliable make-models as well as for faster deteriorating make-models.