(Burtless and Moffitt, 1984; Börsch-Supan and Schnabel, 1998, 1999). As a result, data with which to study retirement will need to include a wide range of information on potential income sources and earning opportunities (National Research Council, 1997; Hurd, 1998). Moreover, the collection of these data must begin much earlier than at normal retirement age, possibly as early as age 45. Health status also plays a role in retirement decisions (Bound et al., 1999), and it will therefore be necessary as well to stress the importance of accurate health and disability information. It may well be that changes in workplace environment or even the act of retirement itself may affect health and disability outcomes. Longitudinal data that record the timing of these work, retirement, and health events are therefore essential.
A further complication arises from the need to compute accurate summary measures of retirement incentives. As discussed below, different countries and different individuals within a country vary considerably with regard to the mix of private and public pension entitlements. Moreover, contribution periods and contribution rules can pose significant challenges for lifetime data collection. However, relatively simple summary measures can be collected through a sensible mix of longitudinal survey data and access to administrative sources. Indeed, the need to supplement standard sample survey data with administrative data is a major requirement for accurate analysis and inference, a theme to which we return in Chapter 8.
Measuring the work opportunities for older individuals requires careful consideration of the changing demand for older workers (Straka, 1992; Organization for Economic Cooperation and Development, 1995b). For example, the earnings profiles of those who remain in work may provide a misleading guide to the earnings opportunities of those who leave the labor force early. This will be especially true if those who withdraw early are predominantly from a single skill group or industry. For measuring changes in work opportunities there is no real alternative to having accurate earnings histories with skill and occupational information.
Comparative analysis is particularly effective in enhancing understanding of the issues surrounding work and retirement, and there already exist many successful comparative studies in this area (see, e.g., Quinn and Burkhauser, 1994; Gruber and Wise, 1998, 1999; Johnson, 1999; Hermalin and Chan, 2000; Borsch-Supan, 2000b; Disney and Johnson, in press). These studies have documented the differing experiences of various countries and have significantly enhanced understanding of the retirement process. Comparisons are particularly instructive among countries that have experienced differing trends in retirement but, on the basis of standard economic and demographic statistics, might otherwise appear rather similar.