ment earnings. This definition of the “Social Security replacement rate for individual earnings” is widely used, not just currently by the Social Security Administration, but by others (see, e.g., Congressional Budget Office, 2001:20-21). However, these “replacement rate” figures may be—but aren’t always—markedly lower than those associated with other definitions of replacement rates, such as that used in most retirement planning, with the same or similar economic circumstances. Steuerle et al. (2000) warn against confusing the two definitions in policy discussions.2
Simply stated, the Social Security replacement rate as defined in this report compares that illustrative worker’s benefits on retirement with her or his preretirement earnings (below the maximum relevant for taxation and benefits), throughout her or his career, indexed by average earnings throughout the economy. In contrast, the definition of replacement rate used in most retirement planning compares a family’s income in retirement from all sources (e.g., Social Security benefits plus any labor earnings, income from pensions and income from private savings), to the family’s income from all sources, typically in the years immediately preceding retirement.
There are two basic distinctions between the two definitions of retirement rates. First, the Social Security definition understandably is limited to Social Security benefits, which derive only from covered payroll earnings. The retirement definition is broader—including all sources of income.
The second distinction between the two definitions is the simplest explanation of why the Social Security replacement rate often, but not always, is markedly lower than the measure used in retirement planning. The Social Security definition is for individual workers—not couples or other family groups—and thus excludes the Social Security benefits any spouse receives in retirement; it also excludes any spousal preretirement earnings. Because spousal Social Security benefits are often considerable, because there are now many two-earner couples (and likely to be in the future); because spouses’ labor force attachment and rewards increasingly approximate each other; and especially because most people enter retirement married, the Social Security replacement rate is generally lower than that used in retirement planning.3
The research on this topic calculates replacement rates from rich data sources, rather than solely the program’s administrative records: Lusardi and Mitchell (2007) and Mitchell and Phillips (2006, 2009) use data from the Health and Retirement Study; Biggs and Springstead (2007) use the modeling income in the near-term microsimulation model of the Social Security Administration, also with longitudinal microdata at both the family and individual levels. The last study focuses on one aspect of the question: the degree to which benefits and other income sources in retirement replace preretirement earnings, rather than replacing preretirement income from all sources.