The effect of privacy protection on survey data accuracy is not well understood, es­pecially relative to other sources of survey error. A common evaluation approach is to compare statistics generated from the survey with versus without privacy protection. However, this approach implicitly assumes that the survey before privacy protection is the “truth,” which we know is not often the case. When there is already error in the survey, differences between the original and privacy-protected data do not directly translate to accuracy reductions. We demonstrate an improvement to this approach by extending the total survey error framework to include error from privacy protection and applying it to linked survey-administrative data with and without synthesis ap­plied to the survey. Doing so allows us to evaluate the relative magnitude of coverage error, measurement error, item non-response error, and privacy protection error. We also evaluate the differential effect of these errors on demographic sub-groups, which impacts frequently used measures of inequality. Using American Community Survey data linked to administrative and proprietary data, we find that error from privacy protection in a select set of estimated means and population sizes is smaller on average than measurement error or coverage error and similar in magnitude to non-response error. Error from privacy protection tends to be larger for categorical variables than continuous variables, while the opposite is true for measurement error and for non­response error. Additionally, error from privacy protection tends to shrink estimated outcome gaps between demographic sub-groups, whereas measurement error tends to inflate those same gaps.