To what extent do the outcomes of markets and other social institutions emerge from complex micro-level interactions within the constraints of their structures, as opposed to being constructed from behavior of individual agents populating them? We present results of a computational experiment in which four important rules of laboratory continuous double auctions (CDAs) are relaxed by introducing white noise in their implementation: (1) no-loss constraint on bids and offers, (2) price priority among bids and asks, (3) bid-ask matching, and (4) trade order of multiple endowed units. Market level outcomes--allocative efficiency, prices, trading volume, and root mean squared deviation of transaction from equilibrium prices--in CDAs are initially robust to introduction of noise. Erosion of these outcomes (relative to predictions of Walrasian equilibrium) is gradual with the rise in noise levels and becomes significant at relatively high levels of noise. Noisy implementation of some rules is more disruptive to outcomes. This robustness of market outcomes may help explain the tendency of certain markets to yield outcomes close to theoretical equilibria derived from assuming traders’ profit-maximization under a broad range of circumstances beyond classical conditions for perfect markets.