"This paper investigates the reasons for discrepancies between the pension plan type reported by respondents to the Health and Retirement Study (HRS) and pension plan type obtained from documents produced by their employers, called Summary Plan Descriptions (SPDs). The analysis suggests the discrepancies are sizable and are mainly due to misreports by respondents. Discrepancies between respondent and firm reports of plan type are first documented for different years and from different data sources. Changes over time in respondent and firm reports are analyzed for those who say their plans did not change. Plan type from payroll data produced by Watson Wyatt, a pension consulting company, is examined and compared to respondent reports for employees covered by Watson Wyatt plans. The Watson Wyatt payroll data report plan type without error, and yet we find the patterns of discrepancies between respondent and firm provided data are the same as for the HRS employer and respondent data. We also explore other evidence gathered by the HRS in the course of interviews and various experiments. Our findings that errors are mainly the result of misreporting by respondents, together with findings from experiments, suggest a number of changes in survey design that can help to reduce reporting error. They also suggest that models of retirement and saving behavior should allow for imperfect knowledge by decision makers"--National Bureau of Economic Research web site.