retirement income and availability and health considerations are the major determinants of labor force withdrawal. Thus, the relative size and age distribution of the labor force in the next century should be strongly influenced by how these two factors develop over time.

B. IMPACT OF AGING ON THE UNEMPLOYMENT RATE

The overall unemployment rate reflects numerous contributing factors, including those related to economic cycles and those which are noncyclical in nature. Recently, attention has been given to the noncyclical determinants of unemployment, with the major determinant in this category being the demographic composition of the population. Studies of the impact of changing demographics on past unemployment rates are useful for the insight they offer with regard to the impact of future demographic changes on unemployment rates. Flaim (1979) analyzed the effect of demographic changes on the overall unemployment rate over the past two decades. He found that, over the period 1957-77, changes in the structure of the population accounted for a 0.6 to 1 percent increase in the overall unemployment rate. This increase was largely due to the entry into the labor force of the baby boom cohorts. When examining the compositional changes which have occurred since 1967, Flaim finds that changes in the demographic composition of the population and changing labor force participation rates have exerted a relatively equal impact on the unemployment rate. Flaim projects that in the near future, the declining proportion of youths in the labor force will exert a downward pressure on the unemployment rate. He predicts that, by the year 1990, changes in the population composition will account for an unemployment rate that is half a percentage point below what it would have been in 1977.

Gordon (1973) estimated that changes in the age-sex distribution of the labor force between 1956 and 1967 caused an increase in the national unemployment rate of about 0.25 percent. Similar calculations for the 1956-70 period show that compositional changes caused an increase of 0.34 percent.

Perry (1970), in examining unemployment rates by age-sex groups, found that, as the relative size of an age-sex group has grown, its relative unemployment rate has worsened. A prime example of this phenomenon can be found in the young age groups. The converse is also true, so that, as the proportion of prime age males has declined, so has that group's unemployment rate. Perry further states that, while substitution in unemployment across age-sex groups has occurred, it has been insufficient to prevent the increasing divergence in age-sex group unemployment rates. For example, workers under age 25 comprised 15 percent of the total employed in 1956 and 20 percent in 1969, yet 31 percent of the total unemployed in 1956 and 50 percent in 1969 were under age 25.

Wachter (1976) also stresses the relatively low substitutability of younger and older workers, stating that older workers are more established in their career paths and have accumulated valuable onthe-job training which younger workers cannot supply. He attributes the significant rise in the nonaccelerating inflationary rate of unem

ployment over the past 15 to 20 years 29 primarily to changing demographics.

Stafford (1979), on the other hand, attributes the longer spells of unemployment which older workers experience to lower rates of accumulation of on-the-job training and stronger employer-specific commitments. Stafford also states that spells of unemployment experienced by older workers often turn out to be "retirement furloughs." Since older workers do have strong employer-specific commitments and do experience longer spells of unemployment, those older workers who are eligible to receive social security or pension benefits may be induced to drop out of the labor force rather than search for jobs.

In conclusion, it appears likely that the declining rate of growth in the labor force (as discussed in the previous section) coupled with the decreasing proportion of the unemployment-prone teenage group will result in lower future aggregate unemployment rates. In addition, there will be greater incentives for employers to retain older workers, hence lowering the unemployment rate and/or decreasing the duration of spells of unemployment for that group.

C. IMPACT OF AGING ON PRODUCTIVITY

What little literature exists regarding age-specific productivity rates indicates that there is little relationship between age and productivity. A series of reports by the Bureau of Labor Statistics (1956, 1957a, 1957b, 1960, 1964) examined the relationship between job performance and age in the clothing, footwear, and household furniture industries, as well as among office workers and Federal mail sorters. These reports generally found that the variability in productivity within age groups exceeded the variability across age groups. A similar study by Hilary Clay (Clay, 1956) of performance in relation to age at two printing shops reached the same general conclusion. Clemente and Hendricks (1973) have also stated that age is a poor predictor of job performance. A more recent study by Schwab and Heneman (1977) of semiskilled, piece-rate workers found that productivity actually increased with age. However, after controlling for experience, the age-productivity relationship was no longer significant.

As Sammartino points out (Sammartino, 1979), the existing research on productivity suffers from two major flaws. First, none of the studies mentioned examines changes in productivity as an individual ages. What is measured is rates of productivity across age cohorts at a single point in time. Thus, it is unclear whether the observed productivity reflects the impact of age or the impact of differing attributes of different age groups. Second, age cohort productivity measures may be suspect due to the problem of selectivity. Presumably, less productive workers are eliminated from the labor force over time, and the measurement of job performance with relation to age does not control for this factor. Thus, it would appear that the relative lack of recent, systematic literature and the questionable methodology utilized in the existing literature leave the relationship between age and productivity open to question.

The nonaccelerating inflationary rate of unemployment is defined as the unemployment rate at which

With regard to labor force growth and productivity at the aggregate level, Leibenstein (1972) states that aggregate productivity is enhanced with labor force growth since young entrants are better equipped in terms of human capital (i.e., education). Therefore, he would state that the greater the turnover from older workers to younger workers, the faster the increases in productivity growth. However, it might also be stated (Serow, 1976) that the decline in turnover (slower labor force growth) would permit increases in the level of human capital of the new entrants which would in turn increase aggregate productivity. Serow (1976) calculated indices of productivity to the year 2020 after estimating productivity by age. He utilized census data on earnings by age as a function of educational attainment to compute weighted mean earnings by age groups, with the weights equal to educational attainment by age. The cohort with the highest weighted average annual earnings was used as the base and set at 1, and the remaining cohorts were expressed as percentages of this base. The index values derived in this manner were: 20-24, .417; 25-34, .760; 35-44, .978; 45-54, 1.0; and 55-64, .824. (This calculation rests on the assumption that earnings are a direct reflection of productivity.) These indices of productivity were then used with labor force projections based on census series D and E population projections. The resulting indices show marginal increases in aggregate labor productivity over time as a function of the changing age composition of the labor force. As table 4 shows, the population aging which takes place under series E projections yields higher levels of productivity than under series D with its lower proportions of older workers.

TABLE 4.-SEROW'S INDICES OF AGGREGATE LABOR PRODUCTIVITY, SERIES D AND SERIES E PROJECTIONS,

Source: William J. Serow, "Slow Population Growth and the Relative Size and Productivity of the Male Labor Force,'' Atlantic Economic Journal, spring 1976, p. 64.

Serow then takes these indices one step further and estimates the relative values for output/labor, output/capital, and capital/labor ratios for alternative demographic situations, from 1970 to 2020. It is anticipated that, with smaller growth in the labor force and smaller inputs of labor, more capital may be utilized per unit of labor. Greater investment in both physical capital and human capital can be anticipated as the growth in the labor force slows. Thus, labor productivity would be higher under the projected slow-growth population than the faster growing population, with the output-to-labor ratio higher and the output-to-capital ratio lower. This smaller output-to-capital ratio is not surprising since, as the supply of capital rises relative to labor, each successive unit of capital input adds progressively smaller amounts to output. Table 5 displays Serow's estimates. Serow's findings confirm the earlier findings of Phelps (1972) and Spengler (1972).

TABLE 5.-SEROW'S ESTIMATES OF RELATIVE VALUES FOR OUTPUT-LABOR, OUTPUT-CAPITAL, AND CAPITALLABOR RATIOS FOR ALTERNATIVE DEMOGRAPHIC SITUATIONS, 1970-2020

Thus, while the evidence on age-productivity relationships at the individual worker level are inconclusive and do not support a generalized statement with regard to the impact of slowing population growth and population aging on productivity, the literature which addressed aggregate productivity levels is more conclusive. Based on the research conducted by Phelps, Spengler, and Serow, it can be stated that declining population growth rates and concomitant population aging will result in greater capital-labor ratios and increased aggregate productivity.

D. IMPACT ON VERTICAL MOBILITY

The consensus in the literature with regard to vertical mobility is that, under conditions of slowing population growth and population aging, such mobility is impaired. Spengler (1971) states that upward mobility is limited in the same manner that mobility within military establishments is limited in the absence of war. Insofar as seniority determines position and status, the proportion of superior positions allocated to older persons would be relatively high while the allocation to younger persons would be relatively low. The access of non-primeage workers (those under age 40) to higher positions would be much more limited under conditions of slow growth than rapid growth. Wander (1972) agrees with this conclusion, and she hypothesizes that smaller proportions of young workers could result in diminution of wage differentials by age, thus removing incentives for younger workers to improve their skill levels to receive higher wages. Spengler (1971) also states that the decrease in vertical mobility would necessitate some "restructuring of remuneration to diminish inequality in the reward structure at least to the level at which performance and output are not adversely affected."

Keyfitz (1973) states that an increasing population facilitates individual mobility. He estimates an equation for the relationship between individual mobility and population increases to demonstrate the effect of population growth on promotion by age. This equation shows that a change in population growth from 2 percent per annum to

59-559 0 - 80 - 5

zero growth implies a 4%1⁄2 year delay in reaching the middle positions of an average factory or office. In terms of higher proportions of elderly individuals in a stationary population, it is reasonable that retirement decisions and mortality would impact upon the promotion opportunities of those remaining in the work force. However, Keyfitz found that a rapidly increasing population is more than three times as advantageous as a high level of mortality in terms of promotions for the survivors. Resignations (or early retirement) are analogous to levels of mortality, for as Keyfitz points out, it is immaterial to those remaining in the labor force whether those exiting do so as a result of voluntary withdrawal or death. However, mortality, resignations prior to age 65, and population increase all contribute to more rapid job advancement.

Since age and seniority are necessarily highly correlated, increasing proportions of elderly persons will inhibit upward mobility. Browning (1975) argues that early exits from the labor force would increase upward mobility. However, while the encouragement of early retirement policies may have advantageous effects upon labor mobility, it is in direct opposition to the encouragement of prolonged work life to compensate for the declining growth in the labor force and increased dependent/worker ratios.

Browning suggests that multiple careers may be the solution to the mobility problem, arguing that when workers reach an impasse to upward mobility in one career they might switch to another. He recognizes that the severe decline in earnings which normally accompanies such movements (since seniority and experience would be inapplicable in a new career) would have to be lessened in order for this to be a palatable solution. Keyfitz suggests three possible solutions. The first involves increasing the indicators of status, the rationale being that if there are four levels of machine operators rather than two, the opportunity for perceived advancement would be increased. His second solution is elimination of all indicators of status. His final solution is to advance technology to the point where gradations in labor skills are less relevant than gradations in technology. Barring these institutional changes or others, it is likely that upward mobility will be dampened by population aging.

E. IMPACT ON INDUSTRIAL MIX AND OCCUPATIONAL DISTRIBUTION

The main determinant of the impact of population aging on the future industrial mix and occupational distribution will be the change in preferences reflected in demand for final products. If, for example, a more elderly population is predicted to demand relatively more services than manufactured goods, then industries will respond to such demand, and employment growth will be heavier in service industries.

Technological advances are another factor which impact upon industry growth; however, growth of this type may not translate into employment growth. Technological advances will be most disadvantageous in terms of employment for the older worker. Skills obsolescence is hastened by technological change, and older workers are more likely to be affected by skill obsolescence. Younger workers with more recent education and training are more likely to be familiar with the latest technological advances. In addition, where training is required, employers are more likely to train younger than older workers,