Integrating a non-probability panel survey with a probability based survey using statistical matching is a promising model for leveraging the statistical benefits of a rigorous probability sample with the cost and timeliness benefits of non-probability web. Statistical matching, or data fusion, is the process of combining data from two separate sources by linking members with similar characteristics. The traditional statistical matching challenge is to join two disparate sets of variables (Y, Z) in order to create a synthetic record jointly measuring Y and Z. In a nonprobability to probability matching application, the focus may be less on joint distributions and more on weighting. The probability sample, which represents the population, provides the distribution to calibrate the nonprobability sample. Each person selected in the probability sample is assigned a statistical match from the nonprobability sample and inherits the nonprobability data from that match. The questions we explore in this presentation is—does this statistical matching approach preserve the sampling properties of the probability sampling, allowing the calculation of sampling errors and subsequently inferential statistics? Are there other forms of variability that must be accounted for in order to capture the total variance? Is the variability from the web panel a good estimate of the population variance? Our research uses data from the National Health Interview Survey, the National Alcohol Survey (dual-frame RDD) and a non-probability web survey.