Comparing the Utility and Disclosure Risk of Synthetic Data with Samples of Microdata