Knowledge that Transforms

B. T. McCallum, The Role of Speculation in the Canadian Forward Exchange Market: Some Estimates Assuming Rational Expectations, The Review of Economics and Statistics, Vol. 59, No. 2 (May, 1977), pp. 145-151

THE monocentric urban area is without doubt the best known theoretical construct in the literature of urban economics. Many variations exist, but all share certain basic features: a city is located about a node on an otherwise featureless plain. This node, the Central Business District (CBD), is the only concentrated place-or the only place-of employment and shopping. Workers live in housing located outside the CBD and travel to and from their workplaces a fixed number of times each day over a uniformly available, radial transportation system. Transportation costs increase with distance of the residence from the CBD and may include either or both money and time. Households purchase their housing and other commodities to maximize a utility function, not including transportation as an input, subject to a budget constraint net of transportation expenses. As is well known, a number of propositions can be derived from these principal assumptions and some subsidiaries; for example, the existence of downward sloping land rent and housing price surfaces and the spatial segregation of households by income class and preferences (Muth, 1969). What is possibly as well known as this basic model is the difficulty of making the assumptions more realistic. As Beckmann (1974, p. 99) has noted, even minor variations tend to run into a thicket of mathematical difficulties. The effects of multiple workplaces, restrictive zoning, local amenities, and racial discrimination are only a few of the features of real urban areas that can be incorporated into the basic model only with great difficulty or not at all. Since analytical solutions to more realistic models have been difficult or impossible, there has been some recent effort to use computer simulation to examine their properties. Mills, for example, has examined the characteristics of a complex model both with congestion costs and with an efficiently priced transportation system, finding the city's land area in the latter case to be greater than in the former. This result, he suggested, would surprise some persons (1972, ch. 8). Hartwick (1974) has used the same model to study the conditions for spatial segregation or integration of various productive activities, asking, in effect, in what circumstances will persons commute to distinct work areas American-style rather than live above local shops European-style. Zeller (1971) has devised a computer simulation model to portray the bid-rent curves of firms making intrametropolitan location decisions. A particularly interesting possibility has become available recently from the work of various persons, notably Scarf (1973), on the computation of equilibrium prices for competitive economies. The algorithms developed are capable of calculating equilibrium prices, demands and supplies for an economy given only initial resource endowments and demand and supply functions. No more than quite modest restrictions need be imposed on the functions, and the characterization of the economy by number of commodities, types of consumers, and interrelationships in production can be quite rich. Since the characteristics of the various urban models are simply those of competitive economic systems in equilibrium, there would seem to be an obvious opportunity to explore the properties of quite complex models with a very general technique. As might be expected, however, efforts to apply the algorithms to models of spatial economies encounter their own problems, both technical and conceptual. To my knowledge, the first successful application to spatial Received for publication October 3, 1975. Revision accepted for publication September 14, 1976. * An earlier version of this article was presented at the European Conference on Housing Markets, Mons, Belgium, June 1976. I am indebted to Curtis Harris for careful reading and perceptive comments on earlier drafts. Herbert Scarf, Barbara Bergmann, Charles Clotfelter, Charles Lieberman, Michaei Murray, and others have also made helpful comments. Computer funds were provided in a faculty research grant from the University of Maryland Computer Science Center. Copies of the algorithm described in this paper are available upon request.

A decade has passed since the passage of the federal Elementary and Secondary Education Act (ESEA) of 1965. Current federal expenditures for compensatory education under that act now exceed $1.5 billion annually. Those expenditures are largely focused on children in the elementary grades, in particular, kindergarten through grade three.' A major objective of compensatory education programs can be viewed as reducing poverty by increasing the human capital stocks of students at the end of their school-lives (i.e., the terminal human capital stock). Proponents of early childhood intervention programs expect terminal human capital stocks to be higher when limited resources are reallocated from later to earlier grade levels. However, empirical research has by and large failed to fulfill this expectation. Evaluations of early childhood intervention programs like Headstart have generally produced inconclusive results (Cicirelli, 1969; Bronfenbrenner, 1974; Ryan, 1974).2 There are several possible explanations for this finding, but the one explored in this research is low marginal productivity, with respect to the stock of terminal capital, of school inputs received in the early grades. While the conventional belief among psychologists and educators has been that this productivity is high (Hunt, 1961; Bloom, 1964), others have argued that the marginal product of school inputs in producing terminal capital is higher in the adolescent years than the early childhood years (Rohwer, 1971). If the latter view is correct, compensatory education programs should maximize terminal capital by increasing school resources in the secondary grades, not the early elementary grades. The purpose of this article is to estimate the productivity of school inputs over time for both and children. The estimates obtained will have implications for the optimal allocation of limited school resources over the school-life.3 In the case of disadvantaged children, the estimates provide one possible explanation for the presumed failure of compensatory, early childhood education programs. In the case of advantaged children, the results offer one prediction of the success of non-compensatory, early childhood education, which is receiving increasing political support these days. The model of human capital accumulation is derived and described in the next section, followed by a discussion of the data and sample used in the estimation of the model. Subsequently, the estimated results are presented, and the policy implications of the results are explored.

Recent studies at both the theoretical and empirical levels (Feldstein and Rothschild, 1974, Nickell, 1975, Feldstein and Foot, 1971, Eisner, 1972 and Bitros and Kelejian, 1974) have offered accumulating evidence inconsistent with the neoclassical investment theory assumption that replacement investment is a constant fraction of the capital stock.' While this assumption was accepted largely on theoretical grounds, with renewal theory implying in the long run (given capital growing at a constant rate) that replacement investment will approach a constant proportion of capital stock, Feldstein and Rothschild have recently presented a series of contrasting theoretical arguments that suggest that it is likely to be untenable. To date, the most conclusive set of empirical results in support of this view has been Bitros and Kelejian's (hereafter B-K) analysis using annual data for the U.S. electric utility industry for the period 1946 to 1971. Unfortunately, the B-K analysis is subject to several important problems which involve the measurement of capacity and the nature of the electric utility industry.2 The purpose of this paper is to re-examine the B-K results, accounting for each of these issues. Given the importance of the proportionality assumption of neoclassical investment theory (particularly in the development of capital stock and user cost series), a reconsideration of the B-K evidence is warranted. Section II briefly reviews the B-K model, data, and results. The third section outlines each of the problems with their analysis. In section IV we present estimates of an amended version of their model using a relatively homogeneous component of the electric power industry, class A and B privately-owned firms for the period 1946 to 1971. Furthermore, we test this model for specifications errors and compare our findings with those of B-K. The last section summarizes the results.

Previous efforts to examine the relationship between sex and student performance in college level introductory economics have generally found that males seem to perform significantly better on final examinations than females. Bach and Saunders (1965), Attiyeh, Bach and Lumsden (1971), Attiyeh and Lumsden (1971), Saunders (1971), Bolch and Fels (1974), and Clauretie and Johnson (1975) all found empirical support for the superior performance of males. The results are consistent across a variety of test instruments, including TUCE, TEU, TEC and CLEP.' Bolch and Fels (1974) have explained why contrary evidence from Buckles and McMahon (1971), Danielson and Stauffer (1972), and Lewis and Dahl (1972) is not convincing. The standard hypothesis underlying these empirical tests is that female students have grown up in a cultural environment in which girls are not supposed to like business and thus have a disadvantage in business or economics courses. More recently Garron (1970) has argued that the difference between male and female performance in learning spatial and numerical skills, which are related to understanding economics, is chromosome linked. This analysis tests these hypotheses once again, using a sample of 153 students from Vanderbilt University in Spring 1975. These new tests have advantages over those previously performed.' First, we employ a test instrument which is targeted at the course material scheduled to be taught in the experimental course. A well targeted exam is desirable because tests that include material with which students are unfamiliar are apt to measure intelligence rather than learning in the introductory economics course. Second, we examine the hypothesis separately for students who took the course in a conventional lecture-discussion type format and students who elected a personalized, self-paced method of instruction (PSI) (see Fels, 1974). Third, previous studies have typically used a final examination score as the indicator of performance. Final exam scores measure the level of understanding and comprehension of economics principles at the end of the course, but do not necessarily accurately reflect learning that took place during the instructional period. Our improved measure of learning is a construct, namely the percentage of the gap between a pre-test score3 and one hundred percent that is closed by a student during the semester.4 This value added type measure is more appropriate for testing hypotheses regarding the relationship between sex and learning during the introductory economics course.5 The results of multiple linear regression analysis using both final exam score and the gap closing measure as a dependent variable are reported in table 1. The independent variables are self-explanatory. Scholastic Aptitude Tests, and grade point average are statistically significant factors in determining performance for most of the regression models. This finding is consistent with earlier studies (see Bolch and Fels, 1974). In addition to the standard explantory variables in models of this type we include a binary variable which distinguishes whether either parent worked in a business related occupation on the grounds that such a situation would provide an environment more receptive toward learning business related skills. Concurrent credit hours are included to allow for the possibility that other demands on students' time during the experimental semester might have caused poor performance as students substituted study on other courses for studying economics. Neither of these hypotheses received empirical verification. Using a one-tail test (hypothesizing that males perform better than females), both the pooled and conventional final exam score regressions show a Received for publication October 9, 1975. Revision accepted for publication September 9, 1976. * This paper is part of an evaluation of the personalized self-paced case-method of instructing elementary economics, supported by the Joint Council on Economic Education and Vanderbilt University. The authors benefited from comments on an earlier draft by Ben Bolch, Steve Buckles, Rendigs Fels and Cliff Huang. 'Test of Understanding in College Economics, Test of Economics Understanding, Test of Economics Comprehension, and College Level Examination Program, respectively. 2This examination is a twenty-five question, multiple choice test called the Test of Elementary Economic Analysis (TEEA). It consists of eight recognition and understanding questions and seventeen application questions. The questions cover scarcity and efficiency (20%o), supply and demand (24%9o), national income accounting and analysis (28%7o), money and banking (16%o), and international finance (12%o). The examination was developed at Vanderbilt, pretested for several semesters, and revised in view of item analysis results. 3The pre-test consisted of the same 25 item multiplechoice exam administered on the first day of classes. I Performance = (post-test pre-test)/(100 pre-test). 'This measure was, to our knowledge, first proposed by Whitney (1960). [2471

The following sections are included:IntroductionMethodologyThe Empirical ResultsConclusionAppendixREFERENCES(This abstract was borrowed from another version of this item.)

For many years, public officials and analysts have been acutely aware of a fundamental interaction between Soviet and U.S. decisions on nuclear forces. The force level, technology, and mix that each side considers adequate for its own security depends on the existing (or anticipated) force level, technology, and mix of its adversary. Various models (see Brubaker, 1972; Brito, 1972; Intriligator, 1975; McGuire, 1965, 1967)-some predating the cold war altogether-have been proposed to deduce the implications of mutual interdependence, so as to understand the nature of this Arms Race process. While numerous theoretical explanations exist, the opportunity for testing out the alternate models is necessarily restricted, in that one must wait for sufficient historical data to accumulate. In view of the fact that 15 years of historical data since the introduction of intercontinental missiles in 1960 are now available in an unclassified form,' certain empirical tests to explore the nature of the arms race and of interaction between the United States and the U.S.S.R. appear feasible and desirable. This paper reports on data collection, model formulation, and preliminary regression analysis into this problem. Since rather more subtle formulations and substantially more exhaustive regression runs are required to produce definitive results, the conclusions reached so far must be qualified as tentative. The first purpose of the study is to analyze the historical data, so as (1) to discern whether an interaction process gave rise to the data; (2) to describe the process in quantitative terms to include the stability of Soviet and American behavior and points of equilibrium in their mutual competition; and (3) in some degree to predict its future course. Because of its inherent intuitive appeal and its correspondence with casually observed force-structure decisions in the U.S. context, there is a strong presumption that some such interaction exists. Therefore, another objective of the study is to compare alternative measures of strategic strength to see which measure yields the best, most intuitively plausible quantitative description of this interaction process, on the prior assumption that such a process exists.

The pseudo data approach to the joint production of petroleum refining and chemicals is described as an alternative that avoids the multicollinearity of time series data and allows a complex technology to be characterized in a statistical price possibility frontier. Intended primarily for long-range analysis, the pseudo data method can be used as a source of elasticity estimate for policy analysis. 19 references.

O UR understanding of the serviceproducing sector of the economy is seriously constrained by the inadequacy of measures of real output for the major service industries. Analyses of industry growth and productivity are only as good as the industry output measures on which they are based; and for many service industries (especially finance, insurance, government administration, health services and education) the real output measures that are generally adopted are poor indeed. In some cases production in the service sector, as measured in the national accounts, is no more than an index of labor input, with the result that the calculation of productivity change is essentially a tautological exercise. In almost all cases economists have had to reconcile themselves to the fact that at least part of the disparity in productivity growth between the service and goods-producing industries is a statistical illusion resulting from the inadequacy of existing data and techniques of measurement.