This paper assesses the feasibility of collecting retrospective data on family income in a two-year periodicity in a survey. We draw data and experience from the Panel Study of Income Dynamics (PSID) conducted in the University of Michigan. In 1997, partly because of budget constraints and partly because of an interest in extending the PSID into new content areas, the study switched to biennial data collection from the annual data collection carried out since the beginning of the survey in 1968. After the change, there was a great interest in preserving a few core economic measures for the calendar year (CY) two years prior to the survey year (SY) in addition to the traditional "prior CY" income data. The CY data two years prior have come to be referred to as 't-2' measures and the measures from one year prior have come to be referred to as 't-1' measures. Since 1999, five waves of 't-2' income data have been collected. This paper documents the quality of these 't-2' data in order to evaluate the feasibility of colleting household income data two years later in a survey. Our analyses provide comparisons of income distributions and annual changes in income during the period when PSID interview was conducted annually versus when the PSID interview was collected every other year. We examine total family income, family head's and spouse's labor income, family asset income, and several other components of the transfer income including public assistance, SSI (Supplementary Security Income), Social Security, and income from other sources such as Veteran Administrations, unemployment compensation, worker's compensation, and pension. We discuss situations in which 't-2' income data may provide reasonably accurate estimates and situations in which there may be large reporting errors