The role of employer-provided pensions and Social Security in shaping employees' retirement and saving behavior has attracted an enormous amount of attention from both researchers and policymakers over the past twenty-five years.1 In the research literature, the almost universal assumption is that workers are fully informed about the rules governing their employer- and government-provided pensions. However, to the limited extent that researchers have been able to test that assumption, results suggest that workers are less than fully informed (Bernheim 1988, Mitchell 1988, Gustman and Steinmeier 1989) and that providing information can affect their behavior (Clark and Schieber 1998; Bayer, Bernheim, and Scholz 1996; Bernheim and Garrett 1996; Madrian and Shea 2000). Despite a general lack of research on the role of information in retirement planning, policymakers have made information a central issue. The Social Security Administration (SSA), for example, recently began to mail workers statements of their accrued and projected benefits to improve their ability to plan for retirement; the SSA also has made a retirement planner available on its website. The U.S. Department of Labor has initiated several programs to examine the extent of workers' lack of information and to improve their knowledge of pensions and of saving for retirement in general. Following the 1997 Savings Are Vital to Everyone's Retirement (SAVER) Act, the National Summit on Retirement Savings, held in 1998, emphasized the need to educate the public about retirement planning through media campaigns and other means. In 2000 the Department of Labor celebrated the fifth anniversary of its Retirement Savings Education Campaign. And recent legislative proposals by Representative John Boehner of Ohio (HR 4747, 4748, and 4749) would significantly expand the scope of investment advice that employers are permitted to offer their employees. This chapter provides a comprehensive analysis of what workers (don't) know about their pensions and Social Security. This analysis is based on information from the Health and Retirement Study, described below. Relative to previous findings and current policy issues, the paper provides five key sets of information. First, it uses more recent data than previous studies. This is important because of the significant changes in the pension universe and in Social Security that have occurred over the last fifteen years. Second, the paper focuses on the distribution of differences between respondents' reports of requested information and linked reports obtained from records provided by the Social Security Administration or from detailed pension plan descriptions obtained from firms, examining the patterns of discrepancies at the level of the individual respondent more than did previous studies. Third, the paper examines the effects of poor information on economic behavior in order to assess the potential benefits of providing better information. Fourth, to improve understanding of misreporting and to provide a foundation for imputing pension and Social Security outcomes when data are not available, the analysis explores whether the differences between the cases that have linked Social Security and pension records and those that do not are related to demographic or other measures. Fifth, the appendix provides information of use to researchers, including an analysis of the relation between respondent-reported earnings histories and linked earnings histories from Social Security records. It also includes a set of equations that researchers can use to impute pension characteristics and plan values for cases without employer-provided pension plan descriptions and for researchers who do not have access to linked pension data. Our findings suggest that workers approaching retirement possess a great deal of misinformation about their pensions. Half of respondents with linked pension data correctly identified their plan type, but fewer than half could identify, within one year, the dates of their eligibility for early and normal retirement benefits. According to the firm-provided data, two-thirds of respondents would be eligible to retire by the time they reached age 55; however, less than half of respondents were aware that they were eligible. Those who were within three years of retiring forecast somewhat more accurately but did not do a much better job of forecasting their age of eligibility for early retirement than the sample as a whole. Eighty percent of respondents with a defined benefit plan either did not think that they were eligible for early retirement or did not know the benefit reduction rate for their plan. Respondents did better in reporting the value of their pension than their age of eligibility, but the unexplained variation is still considerable. Only half of the respondents ventured to guess their expected Social Security benefits, and only half of those came within $1,500 of the actual annual amount. On the whole, respondents were somewhat pessimistic in evaluating their defined benefit pensions, in contrast to findings from earlier studies. Respondents' and firms' calculations of pension benefit amounts were in rough agreement in only 40 percent of the cases. A preliminary analysis of how knowledge of Social Security and pension benefits affects retirement expectations, realization of those expectations, and wealth accumulation reveals complex relationships. Because it is easier to adjust saving downward than upward as one approaches retirement, even symmetric errors in expectations should affect retirement and saving outcomes. Yet respondents' lack of knowledge about Social Security and pension wealth and their inability to identify their plan type had only modest effects on retirement plans, on whether those plans were met, and on saving outcomes. Although researchers would like to work with the true value of pensions and Social Security benefits, in many surveys only respondents' reports were available. Our findings show that respondents' reports and other information about the respondents accounted for 80 percent of the variation in linked employerreported pension values and that respondent-reported work histories and other explanatory variables accounted for 75 percent of the variation in earnings obtained from linked Social Security records. Thus prospects are good for imputing pension and Social Security values, although they are not good for imputing the timing or size of incentives for early retirement. Implications for policy depend to an important degree on two considerations: the precise behavioral channels through which misinformation affects retirement and saving and whether increased educational efforts affect behavior and planning in a timely manner. However, there is little information on which to base an answer to either question.