This article shows how state dependence effects can be estimated for many markets and with few assumptions by using data on how the shares buying specific products differ between those who bought the same product on their latest purchase occasion and other consumers. Using as instrument information regarding which product was cheapest when consumers made their last purchase, I estimate that state dependence increases the probability that consumers will buy the product they bought the last time by 8 percentage points. This effect is larger for women and the elderly than for men and younger consumers. The state dependence effect is also larger for brand-names than for generic products, but not significantly related to number of previous purchases.