Knowledge that Transforms

PERHAPS of utmost importance to the longterm structure and performance of the railroad industry are the large-scale corporate mergers which are currently pending or in the planning stage. Though railroad managers and the Congress have apparently been very enthusiastic about the benefits of rail mergers,' this enthusiasm is not shared in the conclusions of retrospective studies on the topic. For instance, Gallamore (1969) and Sloss, Humphrey and Kruttner (1975) concluded that recent mergers have had little success in achieving anticipated cost savings. Unfortunately, these studies do not shed much light on the social desirability of rail reorganization as they fail to measure the benefits which accrue to shippers through mergerinduced improvements in the quality of rail service. In a recent paper, Levin and Weinberg (1979) used post-merger changes in market shares to measure the effect of mergers. They found that end-to-end mergers did increase the market shares of the firms in the sample, whereas parallel mergers did not. In their analysis, increases in market share are assumed to reflect social benefits; however, as acknowledged by the authors2 and demonstrated by Spence (1975), changes in a firm's performance are not likely to be a sufficient criterion for determining the social value of service quality improvements. In general, an analysis of railroad mergers is complicated by the fact that one is analyzing a network industry. That is, performance in the rail industry (and other network industries such as telecommunications, trucking, and electricity) is generally dependent upon operations by a number of distinct firms which jointly produce the final output. In rail freight transportation, for instance, roughly 70% of total car-miles consists of interline service (i.e., involves two or more carriers). Unfortunately, previous analyses of the rail industry have failed to incorporate two critical features of the industry: the network interdependencies among carriers and links in the network, and the effects of differences in service quality upon users of the rail system. In this paper, we explicitly recognize that rail costs and service quality are significantly affected by market competition and/or coordination among rail carriers. From this perspective, we attempt to estimate the potential consequences of rail mergers, vertical or parallel, respectively. We measure separately two classes of potential effects of rail mergers: the cost savings which might be realized by rail carriers (through increased operating efficiencies and improved capacity utilization) and the improvements in service quality which would potentially accrue to the users of the rail system. In the next section, we identify the economies which can be potentially realized through horizontal and vertical mergers. In section III we develop the framework within which the cost and service-related benefits are estimated and describe our data base. The empirical results are presented and interpreted in section IV. Finally, the implications of the results for public policy toward mergers are discussed in the concluding section.

A merican workers have been retiring from I L their jobs and receiving Social Security retirement benefits at younger and younger ages in recent decades.' Motivated by concern over this trend, studies of the retirement decision have followed two distinct lines: some have found poor reported health to be associated with earlier retirement, while others have analyzed the effects of retirement program parameters on life-cycle employment plans.2 There is, however, a connection between these causal factors which is not widely recognized: under the Social Security system, the financial attractiveness of early, actuariallyreduced benefits can be shown to be directly related to the remaining length of time that one expects to live. This is the hypothesis that I propose in this paper. The Social Security system, at its inception, provided retirement benefits only to insured workers who had attained age 65. More recently (since 1956 for women and since 1961 for men), however, reduced benefits have been available beginning at age 62. The reduction is applied to all future benefits, is scaled according to the number of months before age 65 that benefits are first received, and is roughly fair: the reduction is calculated to provide roughly the same present discounted value of lifetime benefits, regardless of the age at which benefits are first taken.3 Superficially, then, the early retirement option appears to be formulated in such a way as to minimize behavioral distortions by providing no financial advantage, relative to taking benefits at 65. Actuarial fairness also makes the option seem costless to the Social Security trust funds. In fact, however, the actuarial reduction equates expected present values of early and normal retirement benefit streams only for those with typical life expectancy. Prospective retirees with atypically low survival probabilities can maximize the expected present value of their benefits by choosing early benefit entitlement,4 while those with high probability of survival wan receive greater lifetime benefits through age 65 entitlement. Adverse selection can thereby increase the average expected present value of benefits for the population as a whole. The paper begins by demonstrating that early entitlement provides the greatest discounted lifetime benefit stream if and only if one dies before a specific borderline age. It is then straightforward to incorporate this actuarial insight into a more conventional analysis of the labor-leisure choice made by 62-year-olds. I then test the hypothesis that age at entitlement is positively related to longevity by using Social Security's Continuous Work History Sample (CWHS). The empirical approach is to measure how well entitlement decisions predict mortality. By taking this approach, I am able to obtain an estimated equation with which age at entitlement can be used to predict actual mortality. Thus, mortality rates can be generated for those claiming benefits at different ages, and by using these it is possible to estimate the gains to beneficiaries (and cost to the Social Security system) due to adverse selection. Received for publication May 3, 1982. Revision accepted for publication January 31, 1983. * Michigan State University. I would like to thank Joseph Applebaum, Paul Cullinan, Robert Dalrymple, Daniel Hamermesh, Kathy Ruffing, Frank Sammartino, and two anonymous referees for advice and comments on earlier drafts. I have also benefited from ideas expressed by Francisco Bayo and Steven Goss of the Office of the Actuary of the Social Security Administration. All errors and omissions are my own, however. Anne Rader provided extremely able research assistance. l Between 1971 and 1976 the proportion of retired workers' benefits awarded to 62-year-olds rose from 29% to 37%. See Social Security Administration (1971-76). In this paper the Old Age and Survivors Insurance (OASI) program will be referred to as the Social Security Retirement Program. 2 Boskin (1977) has highlighted this dichotomy in the literature. 3Monthly benefit amounts are reduced by 5/9% for each month before age 65 that they are taken. Thus, a monthly benefit claimed at age 62 is equal to 80% of the benefit available to the same worker at age 65 (if potential benefits are unchanged by any further covered earnings between ages 62 and 65). An increase of much lesser magnitude (1/4%) is applied to the benefits of those who delay their claims beyond age 65. This increase is far less than actuarially fair. 4 Entitlement denotes the receipt of benefits. Initial entitlement need not be coincident with retirement, except in the sense that entitlement requires that a beneficiary be earning little enough that the earnings test does not reduce benefits to zero.

Stephen E. Haynes, Joe A. Stone, Secular and Cyclical Responses of U.S. Trade to Income: An Evaluation of Traditional Models, The Review of Economics and Statistics, Vol. 65, No. 1 (Feb., 1983), pp. 87-95

Interregional differences in average wages and earnings have been observed particularly in the North and South of the United States ever since the mid-1800s. That observation has motivated several empirical attempts to determine the source of those differentials, measured both in nominal and real terms, and to explain why they have been maintained over time. The general conclusion reached by the overwhelming majority of these studies is that the labor market has not eliminated these wage differentials even in the face of substantial interregional migration. This result has at least two alternative interpretations. First, it would appear to contradict the theory of compensating differences as applied to the labor market (Thaler and Rosen, 1975), which stresses that under the assumptions of perfect information, free geographic and intersectoral labor mobility, and homogeneous consumer tastes, the nominal wage rates of workers who have similar human capital characteristics, live and work in similar environments and experience similar living costs, are driven to equality. Second, this result may only reflect an aggregation error. In other words, there may be several types of labor that are each paid different equilibrium wage rates and comprise different percentages of the workforce in each region. Even if the real wage paid to each class of workers is interregionally invariant, a situation that instead would support the theory of compensating differences, failure to distinguish accurately between labor types could produce the illusion of a wage differential. This paper considers the two alternative interpretations given above as to why interregional wage differentials might exist. Hedonic real wage equations are estimated for four regions of the United States using observations on individual household heads drawn from the 1976 Panel Study in Income Dynamics (PSID). This sample is of interest because the 1976 PSID data contain unusually detailed measures of education, work experience and occupation, as well as information on workplace and job characteristics. Thus, a more complete specification of the wage equation is permitted and the possibility of aggregation error is reduced, particularly in comparison with other interregional wage differential studies. Several of these studies, for example, have been based on aggregate data from the Census of Manufactures (Fuchs and Perlman, 1960; Gallaway, 1963; Scully, 1969; and Coelho and Ghali, 1971) which provide no direct measurements on the human capital of workers. The remainder of the discussion is organized into three sections. Section II specifies the wage equation and describes the PSID data. Section III, then, reports empirical results which are consistent with the findings, based on aggregate data, of Bellante (1979) and Coelho and Ghali (1971) in that they support the theory of compensating differences. More specifically, for full-time workers, the rewards to attributes relevant in determining real wages apparently are interregionally invariant. However, because this result conflicts with most previous research on interregional wage differentials based on aggregate data and virtually all such research based on microdata (Welch, 1966; Hanoch, 1967; Hanushek, 1973, 1981; Hirsch, 1978; and Sahling and Smith, 1983), a number of empirical comparisons are made between the present study and the approaches taken by other investigators. Conclusions and implications are drawn out in section IV.

This paper reports on another attempt to statistically uncover evidence of embodied technological change as an explanation of changes in labor productivity in the United States. In this version of the test, an interregional cross-section, time-series sample of data was used. While the hypothesis has a lot of appeal, it has proven difficult to find manifestations of embodiment in econometric tests. This study has proven to be no exception. Despite the unique data set, results were again not supportive of the hypothesis. A possible limitation of this test is that the time period studied may have been too short. REFERENCES

No abstract provided.