PubMed2000
In order to assess the effect of Social Security reform on current and future workers, it is essential to accurately characterize the initial situations of representative workers affected by reform. For the purpose of analyzing typical reforms, the most important characteristic of a worker is the level and pattern of his or her preretirement earnings. Under the current system, pensions are determined largely by the level of the workers' earnings averaged over their work life. However, several reform proposals would create individual retirement accounts for which the pension would depend on the investment accumulation within the account. Thus, the pension would also depend on the timing of the contributions into the account and hence on the exact shape of the worker's lifetime earnings profile. Most analysis of the distributional impact of reform has focused, however, on calculating benefit changes among a handful of hypothetical workers whose relative earnings are constant over their work life. The earnings levels are not necessarily chosen to represent the situations of workers who have typical or truly representative earnings patterns. Consequently, the results of such analysis can be misleading, especially if reform involves introducing a fundamentally new kind of pension formula. This article presents two broad approaches to creating representative earnings profiles for policy evaluation. First, we use standard econometric methods to predict future earnings for a representative sample of workers drawn from the Survey of Income and Program Participation (SIPP). Our statistical estimates are based on a simple representation of typical career earnings paths and a fixed-effect statistical specification. Because our estimation file contains information on each worker's annual earnings from 1951 through 1996 as reported in the Social Security Administration's earnings files, we have a record (though an incomplete one) of the actual earnings that will be used to determine future benefit payments. Our estimates of the earnings function permit us to make highly differentiated predictions of future earnings for each member of our sample. By combining the historical information on individual earnings with our prediction of future earnings up through the normal retirement age, our first approach produces tens of thousands of predicted career earnings paths that can be used in microsimulation policy analysis. Our second approach to creating lifetime earnings profiles is similar in some ways to the traditional method. For example, it is based on the creation of only a handful of "stylized" career earnings patterns. An important difference with the traditional method, however, is that we define the career earnings patterns so that they are truly representative of patterns observed in the workforce. We use simple mathematical formulas to characterize each stylized earnings pattern, and we then produce estimates of the average path of annual earnings for workers whose career earning path falls within each of the stylized patterns we have defined. Finally, we calculate the percentage of workers in successive birth-year cohorts who have earnings profiles that match each of the stylized earnings patterns. Although this method may seem simple, it allows the analyst to create stylized earnings patterns that are widely varied but still representative of earnings patterns observed among sizable groups of U.S. workers. The effects of policy reforms can then be calculated for workers with each of the stylized earnings patterns. Our analysis of U.S. lifetime earnings patterns and of the impact of selected policy reforms produces a number of findings about past trends in earnings, typical earnings patterns in the population, and the potential impact of reform. The analysis focuses on men and women born between 1931 and 1960. Along with earlier analysts, we find that men earn substantially higher lifetime wages than women and typically attain their peak career earnings at a somewhat earlier age. However, the difference in career earnings patterns between men and women has narrowed dramatically over time. Workers with greater educational attainment earn substantially higher wages than those with less education, and they attain their peak career earnings later in life. For example, among men with the least education, peak earnings are often attained around or even before age 40, whereas many men with substantial postsecondary schooling do not reach their peak career earnings until after 50. Our tabulations of the lifetime earnings profiles of the oldest cohorts (born around 1930) and projections of the earnings of the youngest profiles (born around 1960) imply that the inequality of lifetime earnings has increased noticeably over time. Women in the top one-fifth of female earners and men in the top one-fifth of male earners are predicted to receive a growing multiple of the economy-wide average wage during their career. Women born between 1931 and 1935 who were in the top fifth of female earners had lifetime average earnings that were approximately equal to the average economy-wide wage. In contrast, women born after 1951 who were in the top fifth of earners are predicted to earn almost 50 percent more, that is, roughly 150 percent of the economy-wide average wage. Women with a lower rank in the female earnings distribution will also see gains in their lifetime average earnings, but their gains are predicted to be proportionately much smaller than those of women with a high rank in the distribution. Men with high earnings are also predicted to enjoy substantial gains in their relative lifetime earnings, while men with a lower rank in the earnings distribution will probably see a significant erosion in their typical wages relative to the economy-wide average wage. That is mainly the result of a sharp decline in the relative earnings of low-wage men born after 1950. In creating stylized earnings profiles that are representative of those of significant minorities of U.S. workers, we emphasized three critical elements of the earnings path: the average level of earnings over a worker's career, the upward or downward trend in earnings from the worker's 30s through his or her early 60s, and the "sagging" or "hump-shaped" profile of earnings over the worker's career. That classification scheme yields 27 characteristic patterns of lifetime earnings. Surprisingly, the differnce between men and women within each of those categories is quite modest. The main difference between men and women is in the proportions of workers who fall in each category. Only 14 percent of men born between 1931 and 1940 fall in earnings categories with the lowest one-third of lifetime earnings, whereas 53 percent of women born in those years have low-average-earnings profiles. On the other hand, women born in those years are more likely to have a rising trend in lifetime earnings, while men are more likely to have a declining trend. We find that the distribution of lifetime earnings contains relatively more workers with below-average earnings and relatively fewer with very high earnings than assumed in the Social Security Administration's traditional policy analysis. For example, the "low earner" traditionally assumed by the Office of the Chief Actuary is assigned a level of average lifetime earnings that we find to be higher than the average earnings of persons in the bottom one-third of the lifetime earnings distribution. The stylized earnings profiles developed here can be used for policy evaluation, and the results can be compared with those from the more traditional analysis. That comparison produces several notable findings. Because earnings profiles that are actually representative of the population tend to have lower average earnings than assumed in the traditional analysis, workers typically accumulate somewhat less Social Security wealth than implied in the traditional analysis. On the other hand, because the basic benefit formula is tilted in favor of lower-income workers, the internal rate of return on Social Security contributions is somewhat higher than detected in the traditional analysis. Moreover, the primary insurance amount measured as a percentage of the worker's average indexed earnings tends to be higher than implied by the traditional analysis. Finally, the stylized earnings patterns can be used to compare benefit levels enjoyed by workers under the traditional Social Security formula and under an alternative plan based on individual investment accounts. That comparison shows, as expected, that the traditional formula favors low-wage workers and one-earner couples, while an investment account favors single, high-wage workers. Comparing two workers with the same lifetime average earnings, the traditional formula favors workers with rising earnings profiles (that is, with lifetime earnings heavily concentrated at the end of their career), while investment account pensions favor workers with declining earnings profiles (that is, with earnings concentrated early in their career).