An econometric model of metropolitan development

These areas are Akron, Atlanta, Baltimore, Boston, Buffalo, Chicago, Cincinnati, Cleveland, Columbus, Dallas, Dayton, Denver, Detroit, Ft. Worth, Houston, Indianapolis, Jersey City, Kansas City, Los Angeles-Long Beach, Louisville, Memphis, Miami, Milwaukee, Minneapolis-St. Paul, Newark, New Orleans, Oklahoma City, Philadelphia, Phoenix, Pittsburgh, Portland, Rochester, St. Louis, San Antonio, San Diego, San Francisco-Oakland, Seattle, Tampa-St. Petersburg, and Washington, D. C. New York has been excluded from the sample because of its size and other special characteristics. During the period 1954–1958, manufacturing employment decreased in 24 of the 39 central cities used in this study, wholesaling in 12, retailing in 10, and services in only one. Population measured using constant 1950 boundaries declined in 21 of these areas. For treatment of household location behavior and the choice of transportation modes, see John F. Kain,A Multiple Equation Model of Household Locational and Tripmaking Behavior, The RAND Corporation, RM-3086-FF. April 1962;idem, A Model of Household Location and Tripmaking Behavior with Reference to Detroit, The RAND Corporation, P-2627. The latter was presented at the Ninth International Meeting of TIMS and the Econometric Society, University of Michigan, Ann Arbor, Michigan, September 9, 1962. Basic employment refers to terminology used in economic base studies. “Basic” (primary) employment may be thought of as the exogenous variable in an economic base model. It differs from non-basic (service) employment in that its output is exported and thus provides the area with income. Isard for example, in discussing economic base studies notes that, “These ‘basic’ activities not only provide the means of payment for raw materials, food, and manufactured products which the region cannot produce itself, but also support the ‘service’ activities, which are primarily local in productive scope and market areas.” Walter Isard and Associates,Methods of Regional Analysis: An Introduction to Regional Science, the Technology Press of the Massachusetts Institute of Technology and John Wiley and Sons, New York, 1960, p. 190. The most definitive discussion of Economic Base Models is by Richard B. Andrews, “Mechanics of the Urban Economic Base,” Series and articles inLand Economics, Vols. 29–31, May 1953 to February 1956. When an SMSA has more than one Central City, they have been added together to from a consolidated Central City. This assumption, of course, varies considerably from urban area to urban area. The proportion of manufacturing “basic” employment to total “basic” employment in Detroit and Pittsburgh is many times as great as in Washington, D. C., for example, where Federal Government employment must represent an overwhelming proportion of total “basic” employment. We use manufacturing employment as a proxy for “basic” employment in this model for two major reasons: analysis of inter-metropolitan growth is only of secondary interest to us, our primary concern is the redistribution of activities within urban areas; we have reached the tentative conclusion, elaborated upon at a later point in this paper, that secular shifts in manufacturing employment are the most powerful single determinant of variations in the rate of metropolitan growth. Moreover, there is considerable evidence that manufacturing is a large proportion of “basic” employment for most metropolitan areas. For example, a Kansas City Federal Reserve Bank's economic base study for Wichita indicated that in 1940, manufacturing was 50 per cent of total “basic” employment and in 1950, 71 per cent of “basic” employment. In 1940, the employment categories included in this paper accounted for 83 per cent of “basic” employment and 71 per cent of total Wichita employment. In 1940, the four employment categories accounted for 91 per cent of “basic” employment and 77 per cent of total employment. We are far from satisfied with our treatment of this problem in this paper and hope to be able to obtain greater representation in later work of the type reported here. The Federal Reserve study is reported in Isard,op. cit. p. 191. This is, of course, an oversimplification. Certain types of retailing have had as their major function, service of workplace and out-of-town demand. Our analysis here is limited by data availability. The 1958 Census of Business includes Central Business District retailing employment, but the 1954 and 1948 Censuses of Business did not. However, we have made some provisional estimates of 1954 Central Business District employment in department stores, the largest single Central Business District retailer, and have been engaged in its analysis with what appears to be very promising results. The final model will include both greater spatial detail and greater industry disaggregation. The Municipal Year Book, 1949–59, The International City Managers' Association, Chicago, Illinois. For comparisons of the mean annual percentage changes, the mean annual absolute changes and the number of declining areas, both corrected and uncorrected for annexations during 1948–54 and 1954–1958, see John F. Kain and John H. Niedercorn,Suburbanization of Employment and Population 1948–1975 The RAND Corporation, P-2641, Paper prepared for presentation at the 42nd Annual Meeting of the Highway Research Board, Washington, D. C., January 7–11, 1963. These descriptive statistics for the same 39 SMSAs used in this study are measured for 1950 Central City boundaries. For example see: John H. Niedercorn and John F. Kain,Changes in the Location of Food and General Merchandise Employment within Metropolitan Areas, 1948–58, The RAND Corporation, P-2614. Paper presented at the meeting of the Western Economic Association, Los Angeles, California, August 24, 1962. For a discussion of this hypothesis see Alexander Gerschenkron,A Dollar Index of Soviet Machinery Output, 1927–28 to 1937 The RAND Corporation, R-197, April 6, 1951, pp. 47–48. A recent study by Victor Fuchs on the determinants of manufacturing redistribution provides some insights relevant to this question. Fuchs concludes that, “Between 1929 and 1954 the South and West experienced very large comparative gains in manufacturing employment. This was primarily the result of regional differentials in growth rates of individual industries, not of differences in industrial structure. While the complex interrelation between changes in employment, income, and population do not permit one to draw conclusions with certainty about which was the causal factor in any specific situation, we conclude that ‘demand’ or ‘markets’ has not been the primary determinant of locational change. Industries most clearly recognized as market-oriented did not shift as much as did the others. The most important redistributions occurred in labor-oriented industries such as textiles and apparel, or in industries oriented to natural resources such as chemicals, lumber, and paper. The single most important locational development since 1929, the growth of the aircraft industry in the Southwest, is probably attributable more to climate than to any other factor.” Victor R. Fuchs, “The Determinants of the Redistribution of Manufacturing in the United States Since 1929”,The Review of Economics and Statistics, Vol. XLIV, No. 2, May 1962. An analysis by Richard A. Easterlin using Census of Manufactures data from 1869 to 1947 also provides information pertinent to this question. He concludes that, “The state distribution of manufacturing production became on the whole more similar to that of population in the period. Whereas a redistribution amounting to almost one-third of the manufacturing total would have been necessary to equalize the distribution in 1869, the redistribution necessary in 1947 would have been only slightly over one-fifth.” In discussing the pervasiveness to these declines in a majority of manufacturing industries, Easterlin concludes the widespread character of the declines are due to “factors common to all industries. The most apparent of these is the shift in the location of population and hence of markets.” Richard A. Easterlin, “Redistribution of Manufacturing,” inPopulation Redistribution and Economic Growth United States, 1870–1950, II, Analysis of Economic Change by Simon Kuznets, Miller and Easterlin, American Philosophical Society, Independence Square, Philadelphia, 1960, pp. 108 and 110. The Perloff, Dunn, Lampard and Muth study also bears on the structure of this model; they conclude that “shifts in final markets (as revealed by shifts in population) have been closely associated with major shifts in Manufacturing employment.” H. S. Perloff, E. S. Dunn, Jr., E. E. Lampard, and R. F. Muth,Regions, Resources and Economics Growth (Baltimore, John Hopkins Press, 1960) P. 394. For evidence of this contention in the Chicago SMSA see, Dept. of City Planning, City of Chicago,Industrial Movements and Expansion, 1947–57, City of Chicago and Chicago Metropolitan Area, Economic Base Study Series, Study No. 3, Jan. 1961, especially Map-6, “Metropolitan Location of Manufacturing Establishments which Relocated from the City of Chicago”, 1947–57, p. 31. Income would also theoretically be an important determinant of demand for retail sales. In experimenting with both an independent income and a purchasing power variable, however, income did not appear to make enough difference in the explantory power of the estimating equation or in the statistical reliability of the coefficients to be worth retaining in the model. This may be due to the large errors in our income variable, the size of our geographic aggregates, etc. Now that central city and ring income data are available from both the 1950 and 1960 Censuses of Population, we plan to include again income in our estimating equations. Geographic and industry disaggregation may make a difference in this regard as well. A change in productivity variable might seem appropriate here. However, it does not appear to be helpful in a highly aggregated model like this one. For a rigorous derivation of this model and the use of a change in labor productivity variable in a less aggregate model, see John H. Niedercorn and John F. Kain,Changes in the Location of Food and General Merchandise Store Employment Within Metropolitan Areas, 1948–1958, op. cit. The annexation problem has been one of the most difficult ones involved in using data from the Censuses of Manufactures and Business in this research. These data, as traditionally used in urban analysis, systematically overstate increases or systematically understate losses in Central City employment, thus providing a highly misleading picture of Metropolitan development. Although not perfect, we feel our treatment of the annxation problem has provided satisfactory and reasonably accurate and valid results, Niedercorn and Kain,ibid. For the derivation of the retailing and wholesaling equations refer to the attached Appendix. Finite differences rather than differentials are used in these equations, although in a strict mathematical sense the latter would be appropriate. The selected services variable used here is a veritable baker's dozen including: hotels, motels, tourist courts, i. e. SIC major group 70, except SIC 701 and 7013; personal services, i. e. SIC major group 72; miscellaneous business services, i. e. SIC major group 73; auto repair, i. e. SIC major group number 75; miscellaneous repair, i. e. SIC major group number 76; motion pictures, SIC major group 78; other amusements, SIC major group 79. We plan in future work to disaggregate selected services into two or three more homogeneous groups of industries in terms of their locational propensities. For example, while population is probably a good demand proxy for personal services, various kinds of employment or measures of industry output would very likely be a superior proxy for business services. This diaggregation, we feel, will provide us with much more information about urban structure than do the present relationships. Total United States manufacturing employment grew at the annual rate of 0.15 per cent between 1954 and 1962, but declined at the rate of 0.56 per cent during the period 1954–1958. For an average SMSA with 169,000 manufacturing workers (the mean of the 39 areas), the first growth rate implies an annual increase of 254 employees and the second a decline of 950. The algebraic difference of these two numbers is 1204. Adding this difference to the constance term of Equation 18 gives a constant term of 559 or +0.559 measured in thousands. As previously noted, the vacant land information obtained by questionnaire from city planning commissions includes large errors because of differences in the years in which the surveys were taken, definitional differences, incomplete information and the like. In spite of these difficulties we feel our efforts are amply rewarded by the results obtained and hope in future months to refine and improve these data. These improvements should also provide gains in estimating the model. Of course, some people live in houseboats, but their number is not significant. See Kain and Niedercorn,Suburbanization of Employment and Population, 1948–1975, op. cit. If we consider these omitted employment groups, however, only finance, insurance, some federal and state government employment, and some business services would, on the basis of other information, seem to be groups for which much different locational trends would be expected. Local government employment is nearly as much, or more, population oriented than is retailing, as is a considerable portion of federal and state government employment, i. e. the Post Office, the Social Security Administration, the Department of Motor Vehicles, etc. Construction employment would be located where new building occurs, predominantly in the new fringe areas. Much public utility employment is also population oriented as is much transportation employment. The greater use of motortrucks, piggybacking, the rapid suburbanization of wholesaling and manufacturing, and finally the expansion of air transportation are all trends suggesting a rapid suburbanization of employment in inter-city transporation. Conclusions by the Committee for Economic Development bear on this question, “Certain types of activities, however, show little inclination to deconcentrate. Business and governmental services requiring face-to-face relationships or dependent upon a large pool of female labor continue to exercise strong preference for office space in the core of large metropolitan areas. In eight leading standard metropolitan areas about 80 per cent of all employment in finance, insurance and real estate in 1959 was in the Central Cities. Further evidence of this tendency for office-type activities to concentrate in or near central cities is found in figures on office building construction. From 1946 through 1958 a total of 65 million square feet of new rentable office space went on the market in 24 of the larger cities of the nation.” Committee for Economic Development,Guiding Metropolitan Growth, August, 1960, pp. 17–18. On these conclusions we would comment that these three categories in 1950 represented only about 3 per cent of total employment and that the cities' share of these should not be confused with their rate of deconcentration. Moreover, as the authors themselves point out, 42 per cent of this new office space was constructed in Manhattan. In addition, much of this replaces an older inferior space, and there is evidence that floor space per employee in office activities has increased. Thus very large amounts of new construction do not necessarily imply large employment increases. For example, M. S. Gordon,Employment Expansion and Population Growth, The California Experience, 1900–1950 (Berkeley and Los Angeles, University of California Press, 1954).