Prior laboratory experiments have studied general equilibrium economies constructed from “induced preferences” for artificial goods. We introduce new methods that allow us to study economies constructed instead from subjects' actual, “homegrown" preferences. Our subjects reveal their preferences by choosing portfolios of Arrow securities from budget lines through fixed endowments for a series of prices. We then construct several different economies by sorting subjects according to their revealed preferences. The constructed economies exhibit a wide range of predicted outcomes, where predictions are competitive general equilibria given the revealed preferences. Perhaps surprisingly, in every one of our markets the predicted excess demand is well-behaved, and avoids the pathologies highlighted in the Sonnenschein-Mantel-Debreu theorem. (The main reason seems to be heterogeneity in revealed preferences.) Actual trade in the constructed economies using a tatonnement market institution closely tracks predictions in most markets. The exceptions occur in economies with severe wealth effects that generate excess demands that are at relative to measured preference volatility.