Knowledge that Transforms

T HIS paper reports the results of a study on the import structure of India. To be more precise, the study refers to the territory which was known as British India and covers the time period from 1900 through 1946. The study of aggregate import demand has been done for the two time periods (1) 19001936, a period of 37 years (covering British India inclusive of Burma), and (2) 1900-1946, a period of 47 years (on the basis of the netof-Burma trade data as constructed for the present study). The variable to be explained is net retained private merchandise imports. Gross imports have been adjusted for re-exports and imports on government accounts. Imports in current rupee sums have been deflated by an index of import prices, and a series of net retained merchandise imports in 1939 import rupees has been obtained. Next, the total retained merchandise imports on private account in 1939 import rupees (m) have been disaggregated into three major economic categories: import of consumption goods (mi), import of intermediate goods (mi,,) and import of investment goods (mi). Each subgroup has been deflated by an appropriate price index. The results of this disaggregation, however, conform to the following identity: m 3 me + mi,ft + mi (all in 1939 import rupees). Economic theory teaches us that national income and relative prices are the two major economic variables to be used in explaining variations in the import demand of a country. Unfortunately, there is no official national income series for India for this period. Private researchers have made sporadic efforts to conc+riirt ciir.h cpritc. and la rpcent studvy has been completed for the purpose of constructing a series of national income figures for the entire period of the present study. The overall national income figures would not be expected to give good estimates. Adjustment of the available national income estimates for urban-nonurban effects, or for that matter, construction of a series on urban national income might be helpful. Not many imported goods were consumed in the rural agricultural sector of the Indian economy. Most imports were used either directly for industrial use, or indirectly for the use of the urbanized industrial sector of the economy, the income of which varied positively with the variations in the index of industrial output. Unfortunately, there is no index of industrial output for the period which could be used as a surrogate for national income. Again, there are secondary estimates,2 indices constructed by individual researchers on the basis of overall industrial activity. No index of industrial output as such was available. Activity indices have often given undue weights to financial and other quasi-industrial activities3 which may have no meaningful correlation with the import demand of the country. One approach to this problem is to make use of export earnings of the country as a proxy for national income. Ultimately, each country must be able to pay for imports by export earnings unless it has a perennial capital inflow and/or exports gold as an item of merchandise. In the particular case of India, neither exportation of gold nor such importation of capital was ever a possible course of action. As a matter of fact, over this period, India enjoyed a surplus

T HE last few decades have seen the presentation of a number of new savings theories. Traditionally these theories are grouped under the permanent income, relative income, and absolute income hypotheses. This paper will try to combine ideas from the permanent and absolute theories in order to obtain estimates of the transitory and the permanent income effects.' We will emphasize cross section problems, but our suggestions will also be appropriate to time series analysis. In section II, we will integrate the permanent and absolute savings theories. In section III, we will define normal income as a distributed lag in incomes and use a Koyck transformation to obtain an equation that yields separate estimates of the transitory and permanent (normal) marginal propensities to save. The section will include a discussion of statistical problems in using the Koyck transformation. In section IV, we will use different types of averages to define normal income; then we will derive the relevant savings functions. In the last section, we will present and compare empirical results using all the definitions of normal income.

CONFRONTED by differences in the size of the public sector between political units or within one political unit over time the public finance economist, traditionally, has focused upon the levels of urbanization and industrial development as the major explanatory factors. Often this position is corroborated by relating per capita expenditures to income or wealth as a measure of industrial development. Differences in social, physical, and economic environment, in individual needs, and in the institutional or political setting have not gone unnoticed. Rather, they have been relegated to a secondary position, and scarcely have been related to the levels of consumption of public services. The neglect of these factors may have been valid since they may have had little influence upon differences in the size of the public sector among countries or over great spans of time. This neglect may also have been necessary because of inadequate data and meager quantitative methods. However, the quantitative evidence gathered in this study suggests that when the analysis is limited to the variation in the consumption of public services in major localities in the United States, variables representing all these factors are significant; income is one of the less important.' Method of Analysis

ECONOMISTS have long attacked horizontal merger for its role in increasing concentration and defended it as a relatively painless means of attaining economies of scale and/ or inter-regional entry. This paper attempts to evaluate these divergent views using the experience of six large industries. In part I, the data and definitions used are explained. In part II, an attempt is made to measure the role of merger and of various other factors in concentration change. Part III examines the proportion of acquired capacity that is sub-optimal in scale, and part IV, the proportion that is inter-regional. Part V contains some brief conclusions.

IT is generally believed that foreign capital inflows play a strategic role in promoting progress toward self-sustained growth in developing countries. Yet the relationship between the two has received little rigorous analysis in the voluminous literature on economic development. The Foreign Assistance Act of 1961 enunciated the criterion of as a condition for United States foreign aid,' but the appealing self-help shibboleth has not been given analytical content. However, in a pioneering effort, Professor Rosenstein-Rodan recently raised some of the important analytical questions associated with the self-help notion:

CASUAL observation reveals that plants of different sizes exist and have always existed with little tendency to become more concentrated in any particular size class. This can only mean that plants are not operating under conditions of long-run static equilibrium,' and that concept of scale cannot be used to explain actual size distribution or growth patterns of plants.2 Some writers have suggested idea of an distribution, presumably based on stochastic nature of human ability and human foresight.3 Mathematically, optimal distribution may be related to stationary distribution corresponding to some stochastic process describing growth of plants. This growth in turn is explained by such factors as economies of scale, economies of growth,4 profit margin variations,5 and other dynamic factors. appropriate frame of reference for optimal distribution is expansion path, or scale path. path may be regarded as representing basic manufacturing activity of an industry. Plants cluster around this path, and their sizes are given by their positions on path. size distribution of plants is therefore defined with respect to this path. Similarly, from one period to next, growth of plants is given by their movements along path.6 It is reasonable to assume that this path is characterized by fixed elasticities (rather than fixed proportions) among input and output variables, since plants become more capital intensive as they grow. present paper investigates ( 1 ) extent of economies of scale along expansion path for each of manufacturing industries, and (2) relationship between economies of scale and growth pattern of plants. Our principal hypothesis is that with a given expansion path or given returns to scale all plants tend to expand at same rate. In this case, a strict form of Gibrat's Law would apply and it is possible to speak of an equilibrium lognormal distribution of plants with constant dispersion. When a shift of expansion path takes place, there is also a change in returns to scale along path. Our hypothesis states that there should be a systematic relationship between changes in returns to scale and changes in dispersion of plants. changes in returns to scale result in a differential rate of growth for plants of different sizes, until a new equilibrium lognormal distribution is estab* I would like to thank Frank Child and John Harsanyi for their helpful comments, Joseph King for providing me with empirical materials, and National Science Foundation for financial support. 1 Hymer and Pashigian have argued convincingly that dispersion of firms cannot be attributed to constant returns to scale. Stephen Hymer and Peter Pashigian, Size and Rate of Growth, Journal of Political Economy, LXX (1962), 556-569. In this paper, we are concerned with behavior of plants. However, much of discussion pertaining to size of firm in literature is equally applicable to size of plant. 2Some realism is introduced if Gibrat's Law is made starting point. Gibrat, Hart, Prais and others have discovered that in a large number of industries size distribution of firms is approximately lognormal, and law of proportionate effect that average growth rate is approximately same for firms of different sizes -was generally adopted to explain distribution. Since this results in a continuing divergence of plant distribution over time, it was further proposed that a process of regression is at work, that is, that the firms in any given size class at time t would still be distributed lognormally at t + 1, but their mean size would be nearer to mean size of all firms. P. E. Hart and S. J. Prais, The Analysis of Business Concentration: A Statistical Approach, Journal of Royal Statistical Society, Series A, Part II (1956), 150-191. idea implies that there is some optimal firm size, and plants tend to move towards that optimum. However, it is difficult to argue that optimal size is mean size, because there is little reason for a firm to expand beyond an optimal size. 'See, for example, Milton Friedman's comment on Caleb Smith's paper in Business Concentration and Price Policy (Princeton University Press, 1955). 4 Edith Penrose, Theory of Growth of Firm (New York, 1959). ?Joseph Steindl, Small and Big Business (Oxford Institute of Statistics, Monograph No. 1, 1946). 6 If plant is not on path, its size is determined by position of its projection on path. Of course, technological change and substitution also take place from one period to another, resulting in a shift of expansion path. This is discussed in a later section in paper.

N the postwar period, a great deal has been written about the behavior of the consumer and almost as much (at least in the popular press) on movements in the stock market. This paper presents some estimates of stock market gains and losses of the household sector in the postwar period, and it examines some statistical consumption functions which include the stock market changes as an exogenous variable. The over-all conclusions of this study are that postwar stock market movements, while large and yielding high over-all positive returns on stock holdings, have had little or no impact on aggregate consumption. The major reason for the lack of a statistical relationship with consumption is the fact that the distribution of stock ownership is highly skewed, the bulk of stock resting primarily in the hands of highincome recipients whose spending patterns are little affected by quarterly short-term stock movements.

IN a preceding work, an effort was made to develop a new nation-wide series of the number of dwelling units erected annually in the United States since 1840.1 This series contained revisions of the prevailing estimates for the decade levels of building between 1890 and 1940 and added a full half century to our statistical annals of residential building. decade levels of the estimates up to 1910 were aligned to Census increments of housing stock and nonfarm labor force and were adjusted by rates of new residential building for these increments as found in the state of Ohio. In this paper, the same research effort is extended to estimation of the of nonfarm building, both total and residential. necessary preparatory data regarding the of building in Ohio over the period 1837-1912 has been disclosed elsewhere.2 We seek, first, to establish a reliable and consistent set of decade aggregates of new nonfarm building for the six decades before 1910. For this purpose, we have available periodic Census returns on standing stocks closely related to new nonfarm building activity: (a) the socalled value of nonfarm real estate, and (b) the nonfarm labor force. Periodic readings of the true of real estate will obviously reflect intervening activity in building. These readings will also reflect shifts in urban site values, as gauged by speculative opinion, underlying land improvement, and utilities (roads, sidewalks, water, and sewage, etc.) which become reflected in private land values, and rates of wastage or deterioration of existing stocks. Moreover, true will be variously gauged by analysts of different temperaments with access to different kinds of market information. True measures of reality are thus treacherous. Increments from a historic succession of such measures should have an early unpredictable range of error. We have known, however, from the work of Simon Kuznets, that the Censal surveys of value can be utilized to shed light on building activities.3 These Censal surveys were themselves conducted in the later Census periods with awareness of the difficulty of evaluation. And it is proposed here to use them not in their vulnerable form but in the light of Ohio experience with these same Censal surveys and the independently determined Ohio measures of building activity. A particular Censal survey of market values may have badly misgauged market or the scope of enumeration of property may have been altered, but if the survey did not utilize different standards of evaluation or enumeration in Ohio, the Ohio relationship between building and realty will be affected and the necessary correction can be imparted to the national returns. We have * author is Professor of Economics at the University of Wisconsin-Milwaukee. This paper grew out of intensive research during the past four years into all phases of urban building cycles carried on with the aid of the National Bureau of Economic Research and particularly Dr. Moses Abramovitz. See progress reports on the inquiry in the Annual Reports of the National Bureau of Economic Research for 1962, pp. 48-51 for 1963, pp. 46-47, for 1965, pp. 53-54. An early version of this paper was presented to and discussed at the September 4 and 5, 1963 Conference on Research in Income and Wealth. Conference discussion, particularly the challenging comments of Dr. Paul David of Stanford University, helped to direct attention to contentious issues. 'M. Gottlieb, Estimates of Residential Building, United States I840-I939, (National Bureau of Economic Research, Technical Paper No. 17, 1964). 2 See M. Gottlieb, Building in Ohio Between 1840-1912, Conference on Research in Income and Wealth, Proceedings of Conference September 4 and 5, 1963, Output Employment and Productivity in the United States After 1800 (National Bureau of Economic Research, to be published), 243-290. 3 Simon Kuznets, National Product Since I869 (National Bureau of Economic Research, 1946), 202-215. See the further development of Kuznets' estimates in R. Goldsmith, The Growth of Reproducible Wealth of the United States of America from 1805 to 1950, International Association for Research in Income and Wealth, Income and Wealth Series (1952) 225ff and Appendix Tables. Goldsmith found it feasible to utilize the value estimates of the 1850 Census as the basis for his 1850 wealth estimate (p. 317 and following) while Kuznets worked with the Census returns of 1880 and later.