[Residential Building] [From the U.S. Government Publishing Office, www.gpo.gov] a n 32.314t 1 ? <» 1 K No.1-3 1939_______LllifZ U.S. National resources committee. _________ Housing monograph series. LIBRARY Utah State Agricultural College From .......................... 1 9Q1 « —— . I A w9 Call No.. j R.J l 4:- Acc. No. I O a Q HOUSING MONOGRAPH SERIES, NO. 1 RESIDENTIAL BUILDING BY LOWELL J. CHAWNER A TECHNICAL MONOGRAPH ON ONE PHASE OF HOUSING PREPARED FOR THE INDUSTRIAL COMMITTEE OF THE NATIONAL RESOURCES COMMITTEE The National Resources Committee Assumes No Responsibility for the Views and Opinions Expressed Herein UNITED STATES GOVERNMENT PRINTING OFFICE WASHINGTON! : 1939 ■or sale by the Superintendent of Documents, Washington, D. C. Price 10 cents FOREWORD The general business depression which dominated American life following the boom years of 1926-29 was characterized by a cessation of residential building activities along with the decline of other business. Attempts were made from 1931 on to produce more healthy conditions, so that a revival might come about. Private industry alone, Government alone, and Government and industry together—all have attempted to raise the volume of home construction. All these efforts have resulted in progress but still fall far short of actually producing the number of new houses needed. The National Resources Committee has been thrown in contact with these problems from time to time in connection with various studies.1 During 1936-37 the Committee felt the necessity for securing more detailed information concerning a number of phases of the problems closely related to housing. Technicians in the various branches of the Government were asked to present monographs touching these particular problems. Naturally the men who prepared these documents drew on their own experience and presented their personal views which in no way represent the opinions of the agencies with which they are connected. It was intended originally to try to develop from these individual contributions a comprehensive report on the subject of housing. The natural variety of opinions and approaches to the subject as a whole has made a single report difficult to complete at the present stage of the study. Because, however, certain of the individual contributions are timely in nature and also afford valuable material for technicians in the field, it is felt desirable to make some of them available as soon 1 Cf. Technological Trendsand National Policy, Our Cities—Their Role inthe National Economy, Farm Tenancy, Problems of a Changing Population and Consumer Incomes in the United States, etc. as possible in the form of technical reports. It should be held clearly in mind that these are individual expressions and that the opinions stated are those of the authors and that the National Resources Committee is not responsible for such opinions. The document presented in this instance was prepared by Mr. Lowell J. Chawner, of the Department of Commerce, and deals with some of the broader background factors which influence the demand for housing and the methods of supplying the demand. The statistical method used for presenting the future demand is subject to one major weakness. It attempts to project past trends. While these trends include the achievements of past undertakings, they also include the distortions and aberrations of past building cycles as well as business cycles. In spite of the weaknesses of the method it is stimulating and thought-provoking. Caution should be used in regard to the estimates of future activity: Such activity will result only if the adjustments inherent in the formulae materialize. Furthermore, they are not estimates of need as to number or quality of houses under changed conditions. The numbers of houses required can be materially increased by a stepping up of quality and an increased sense of the need for different types of communities. That changes in the character of demand are upon us seems already apparent. Slum clearance, Greenbelt communities, better design, new ideals for site coverage—all are dynamic in their impact. These are the unknowns in terms of their effects. Our new ideals may confuse our statistical predictions, but they do not change the need for better quantitative understanding of our problems. Because Mr. Chawner’s analysis makes a definite contribution to this quantitative understanding of the problem, his contribution is being made more generally available. October 22, 1938. Mr. Frederic A. Delano, Chairman, Advisory Committee, National Resources Committee, Washington, D. C. Dear Mr. Delano: The Industrial Committee transmits herewith the first of a series of monographs on housing prepared at the request of the President by numerous collaborators from various agencies and assisted by a technical staff. A study in this field was recommended by the Industrial Committee last spring, in connection with investigations of the larger problems of the construction industry. A subcommittee, consisting of Thomas C. Blaisdell, Jr., Lauchlin Currie, and C. R. Chambers (resigned Aug. 1, 1937), outlined the proposal and has given advice in the preparation of the reports under Mr. Blaisdell’s direction. Sincerely yours, Lauchlin Currie Charles W. Eliot, 2d Thomas C. Blaisdell, Jr., Chairman Mordecai Ezekiel Isador Lubin Leon Henderson Gardiner C. Means Harry D. White iii THE RESIDENTIAL BUILDING PROCESS: AN ANALYSIS IN TERMS OF ECONOMIC AND OTHER SOCIAL INFLUENCES By Lowell J. Ghawner 1 The number of families, levels of family income, the cost of competing items of expenditure, and the number of available units influence the price paid for the use of shelter. Rent levels and occupancy on the one hand and building costs, financing costs, taxes, and other costs of ownership on the other hand largely determine the volume of new building in any given year. Several new statistical series measuring these influences, and some fundamental relationships between them are developed in the following section. Introduction The marked fluctuations which have characterized residential building in the United States over the entire period for which reliable measures are available may appear at first glance to be erratic and fortuitous. Fundamentally, however, it is believed that the production of domestic shelter is susceptible to rational analysis in terms of measurable economic and other social influences. Houses, to be sure, differ in several respects from many other commodities, particularly with regard to their pronounced durability. The annual production of houses is thus relatively small when compared with the number of existing structures. Only in a very few years has it been as high as 4 percent of the standing supply even in a country growing as rapidly as was the United States up to recent years. However, as a branch of current industrial activity, residential construction in good years is quite large and has involved the erection of nearly 900,000 family units in nonfarm areas in a single year (1925) at an expenditure of possibly 4% billions of dollars. Single causes are rarely adequate to explain economic processes even for the most rudimentary purposes. In the production and use of domestic shelter, it will be discovered that many varied economic and other social conditions play a highly important part. Marriages and migration, family income, and the competing claims upon income of other items of expenditure as well as building costs and interest rates, site costs and taxes, and similar influences must be carefully appraised in arriving at an understanding of the fluctuations in this industry. 1 Mr. Lowell J. Ghawner is chief of the Division of Economic Research of the Bureau of Foreign and Domestic Commerce, Department of Commerce. The author is greatly indebted to the following members of that Division for Assistance in compiling the statistical series included in this section: For the estimates of increases in the number of families, Esther Wright Staudt; for the estimates of the distribution of families by income groups and the direction of the calculation of the regression equations, Dorothy Smith Coleman; for the estimates of the number of dwelling units annually from 1900 to 1915, Robert Sherman and Harold Wolkind. The statistical materials used in this analysis are stated largely in terms of nonfarm areas. The data necessary in a measurement of the economic factors related to residential building are more satisfactory during the period since 1920 than for earlier years. Some measures are available, however, over the period from 1900 to date. For example, as a part of this investigation there have been compiled beginning with that year a series showing the annual increments in the physical needs for dwelling units in terms of the net increase in families and a series showing the estimated number of units upon which construction was started annually in nonfarm areas in the United States. Analysis of Fluctuations This study of residential building involves two principal stages. First, an analysis is made of the market for shelter, principally from the point of view of the fluctuations in demand. In terms of these demand changes and the changes in the total available supply, an expression is formulated for the price of shelter as measured by rent. Second, an analysis is made of the factors which influence additions to the supply of available units. These factors are outlined broadly in terms of conditions in the market for shelter, measured by rents and vacancies, and conditions influencing the costs of ownership such as purchase price, financing charges, and taxes. As will be noted later, new construction, unless subsidized by public grant or by private philanthropy, tends to occur only when the economic demand for shelter advances to such a point that the return from existing property, either in the form of rental income or of satisfactions to an owner occupant, is in excess of the annual cost of ownership of new units which may be constructed, having equivalent location, facilities, or other conveniences. 1 National Resources Committee 2 The Economic Demand Three principal elements influence the economic demand for shelter: the number of families in a given area, the income of these families, and the competing claims which other items of consumption such as food, clothing, automobiles, and recreation make upon family income. Increase in the Number of Families An increase in the number of families in a given area is not necessarily followed immediately by a period of active construction. Expenditures for new buildings may be postponed several years, and a more intensive use of standing structures always is possible. Such an increasing intensity of use is frequently made in periods of declining income or in periods of increasing costs of ownership. “Doubling-up” in 1933, in excess of that experienced in 1928 or 1929, apparently reduced the occupancy of dwelling units in nonfarm areas in the United States by as many as 500,000 units.2 Notwithstanding these limitations, the rate of increase in families is a fundamental element in the changes in demand. The physical needs for dwelling units in terms of families are essentially local in character, but with suitable allowances for migration may be stated in terms of national totals. It is, consequently, possible to express changes in the physical needs for dwelling units, for example, in nonfarm areas in the United States during a given period, in terms of marriages, plus net immigrant families, minus dissolutions of families by death and divorce, minus customary “doubling-up” of newly married or aged couples with relatives, plus or minus internal migration of families, especially from farm to nonfarm areas. Estimates of the annual increases in families in the United States calculated in this manner are shown in table I. Trends and Distribution of Family Income The economic demand for houses is a function of family income quite as much as it is of the number of families requiring shelter. As a matter of fact, the postponable character of new construction and the ready possibility of doubling or undoubling as the result of moderate changes in income add special importance to trends in income as they relate to building. The character of the market for houses is also greatly influenced by the number of families in the various income groups in different parts of the country or at different periods of time. Figure 1 indicates that the most numerous income classes, including approximately 10,000,000 nonfarm households, had incomes in 1929 of between $1,000 and 2 Statistics on the number of “extra families” are included in the Federal Heal Property Inventory, 1934, U. S. Bureau of Foreign & Domestic Commerce. No measure of the year-to-year changes or of temporary as contrasted with normal or permanent doubling-up is available. Vacancy statistics which are available for a number of cities during the period mentioned are the basis for the estimate indicated above (see fig. 9). $2,000 annually. In 1933, the number of households in these income clashes had declined to some 7,500,000. The number of nonfarm households having incomes greater than $1,000 annually fell from more than 20,000,000 in 1929 to about 12,000,000 in 1933. This shifting of several millions of households into lower income classes greatly reduced the economic demand for shelter and is clearly reflected in the decline in rents over this period. Table I.—Annual net increase in number of families 1 in the United States, 1900-1934 [Thousands] Year Net increase in nonfarm families Net increase in farm families2 Net increase in total families Year Net increase in nonfarm families Net increase in farm families2 Net increase in total families 1900— 239 66 305 1918--.. 269 -17 252 1901 255 66 321 1919—.. 451 20 471 1902 330 65 395 1920 632 -4 628 1903 350 66 416 1921 578 -25 553 1904 295 66 361 1922—.. 598 -86 512 1905 361 65 426 1923 650 -25 625 1906 434 66 500 1924 532 34 566 1907 472 66 538 1925-... 546 -11 535 1908 291 65 356 1926 558 -36 522 1909.. 425 66 491 1927 481 45 526 1910-... 363 60 423 1928 434 50 484 1911 348 52 400 1929.... 490 36 526 1'912 393 42 435 1930 ’275 163 »438 1913.... 451 29 480 1931-.- ’207 165 »372 1914.... 430 13 443 1932.... »91 208 3 299 1915 410 -4 406 1933 »351 57 3 408 1916.... 451 -6 445 1934.... »515 27 ’542 1917.... 486 -35 451 Source: Construction and Real Property Section, Division of Economic Research, Bureau of Foreign and Domestic Commerce. 1 A precise definition of the term family has proven very elusive. The year-to-year changes in the above table are determined by the increments in natural groups such as: man and wife (with or without children and other dependents), and widower, widow or divorcee (with or without dependents). The year-to-year fluctuations in single person “families,” i. e., single individuals occupying a dwelling unit are disclosed only to a very limited extent in the above figures. Since the unadjusted year-to-year changes in marriages less dissolutions, etc., involve substantial assumptions, it was necessary to adjust the year-to-year changes to the decennial increments in private “families” as indicated by the census reports. The census enumerations disclose only the number of housholds, i. e., groups of persons living as an economic unit for the most part boarding together at the same table. However, in view of the fact that all decennial censuses from 1900 to 1930 were taken at periods of fairly comparable economic activity, it is not believed that the above figures involve any considerable trend in doubling or undoubling arising from changes in family incomes. They may, however, reflect some trends in doubling arising from changes in social custom. 2 In some years farm families show a net decrease due to migration of farm families to nonfarm areas in excess of families added in farm areas. The number of farm “families” in 1935 was assumed to equal the number of occupied farm dwellings reported in the 1935 Census oj Agriculture. A number of indications lead to the conclusion that this figure for 1935 may be too high, but no statistical data appear to be available as a basis for arriving at a more nearly correct figure. The net increases for the years 1930 through 1935 were derived in the same manner as those for earlier years. Farm “families” might more properly be designated farm “households.” 3 The total number of families for Jan. 1, 1935 was determined directly from the number as indicated by the annual increments from 1930 to 1935 in marriages, plus net immigrant families, minus dissolution of families by death and divorce. Over the period 1920 to 1930 increments obtained in this manner were 5.3 percent larger than the differences in the census enumerations of private families (with certain corrections to secure comparability). This same correction, 5.3 percent, was applied to the increments just described to give the figures shown in table 1. These figures are considerably larger than the number of independent households added each year during this period due to the doubling up of families as the result of reduced income. In January 1935, the total number of families is estimated to have been 300,000 more than the number of households. Housing Monograph 3 It is important to recognize in any analysis of the housing market that the number of households by income classes should not be related to the new units built but to the total number of existing units corresponding to the total number of households. It is manifestly impossible for economic society to supply all families or the increases in families in all income classes with new units. In nearly all cases, families of low income can be housed more adequately in old but sound units having sufficient space and other facilities for comfortable living than in small and otherwise inadequate structures having the sole advantage Summary of methods: The compilation of this series involved considerable estimation which, however, in nearly all cases was based upon reliable quantitative measures. The data used were: marriage and divorce statistics compiled by the Bureau of the Census for the years prior to 1933 and by Samuel A. Stouffer and Lyle M. Spencer for the years 1933-35 (“Marriage and Divorce in Recent Years,” The Annals of the American Academy of Political and Social Science, November 1936, pp. 56-69); immigration statistics obtained from the files of the Immigration and Naturalization Service; statistics on deaths obtained from the Bureau of the Census; statistics upon farm to city migration obtained from the Bqreau of Agricultural Economics for the years 1929-36; and other materials. The annual increments in the total number of families (farm and nonfarm) were determined first. As noted below, these annual increments were adjusted in each decennial period to equal the ten year increments in private families (more properly “households”) as determined from the census reports. These totals were then broken down to show separately the annual increments in the number of farm and nonfarm families. In arriving at the estimates, essentially the following calculations were made. Marriages.—A considerable number of marriages are contracted by persons who already have homes, particularly widowed or divorced persons who remarry. Based upon the number of widowed and divorced persons who, according to the Census of 1930, were heads of families and the estimated number of such widowed and divorced persons who remarry, it has been estimated that 20 percent of marriages are contracted by persons who already have homes. Consequently, 80 percent of marriages each year were assumed to represent a potential need for new family units. Divorces.—According to the 1930 Census, approximately one third of all divorced persons in the United States are heads of families. Similar data were not collected in other years. Thus, out of every 100 divorces (or 200 divorced persons), it was assumed that 67 “families” remained and that a decrease of 33 “families,” or one-third of the original number of families, occurred. Immigration and Emigration.—All data collected by the Immigration and Naturalization Service pertain to individual persons rather than to families. After studying the available data, it appeared that the best measure of year-to-year immigration or emigration of families was the total number of married females entering or leaving the United States. Throughout the history of the United States until the past few years, immigration has been substantial. For the past three decades, in terms of the above measure, immigrant families have accounted for an increment in families, averaging approximately 50,000 annually until 1924. During 1906 and 1907, this number was more than 100,000 families annually. Since the passage of the Quota Act of 1924, the number of admissible quota immigrants has been restricted to slightly more than 150,000 persons annually. Consequently, under present conditions even with allowances for nonquota admissions, the number of families added from this source is not likely to exceed 30,000 in any year. For the decennial period from 1930 to 1939, there may be a very slight increase by net immigration averaging possibly 5,000 families annually. Deaths.—An estimate of the year-to-year fluctuations in families dissolved by death has been made as a part of this study. This estimate is based upon the approximate percent of married, widowed, and divorced persons of each sex who are heads of families, and an assumption as to the percent to which the deaths of individual persons in each of these groups has resulted in the dissolution of families. The annual number of deaths of each sex in each marital status was estimated for the years 1900 through 1934, using the age specific differentials in mortality between married, widowed, and divorced persons and the total population as computed by Walter F. Willcox, Introduction to the Vital Statistics of the United States, 1900 to 1930, Government Printing Office, Washington, 1933. The decade increments in marriages, plus net immigration, less families dissolved by death and divorce, as outlined above, were in fairly close agreement with the decennial increments as calculated from the reports of the Bureau of the Census (with proper allowances to maintain comparability from decade to decade in the use of the term “families” and in the differences between census dates). The greatest difference, 14 percent, was for the decade 1910 to 1919. Finally, in arriving at the figures shown in the above table, the decade increments in families as reported by the Bureau of the Census, with the adjustments just indicated, were prorated according to the annual increments obtained from the estimates of families as outlined in the immediately preceding paragraphs. INCOME PER HOUSEHOLD DOLLARS Figure 1.—Distribution of the number of nonfarm households by income classes, 1929,1933, and 1935-36. Sources: The distribution of nonfarm “families” by income groups in 1929 was derived from an estimate by the Brookings Institution, America’s Capacity to Consume, table 37, p. 227. As shown on the chart, “families” include unattached individuals operating independent households as well as two or more family groups living together as one household and thus correspond with the census total of “families” or,, more properly, households. The distribution of nonfarm households by income groups in 1933 is based on the percentage distribution reported in the Financial Survey of Urban Housing, United 4 National Resources Committee of being new. The market for existing units, however, clearly bears a close relation to the distribution of households by income classes. Increases or decreases in income and the consequent shifting of households to higher or lower income groups tend to increase or decrease rents in the various classes of property. These shifts in rents, in turn, tend to influence the trends in new building construction. For earlier years, the distribution of households by income groups is not available in as great detail as is shown in figure 1. The average family income each year, however, may be computed and enters into the index of family income shown in figure 2, for the year 1909 and subsequent years. These figures are calculated in terms of actual family groups without regard to changes in doubling and are thus properly described as rates of income per family. Competing Claims of Other Items of Expenditure Improved housing may be considered from time to time more or less desirable by the people of a given community than are automobiles, radios, entertainment, and other items of expenditure. Consumer preferences are not readily measurable, however, in specific terms, and they vary considerably among households at different levels of income. They depend, in part, upon varying prices of commodities and services as well as upon changing social custom. Necessaries of life such as food make strong claims upon income. Increasing costs of such commodities thus in an important degree States Department of Commerce, 1934, and related data from the Statistics of Income, United States Department of the Treasury. The number of nonfarm households in the United States in 1933 was first distributed according to these percentages. (The average number of nonfarm households in 1933 (23,260,000) was estimated using the 1930 census data (see table VII) and the annual increases shown in table I together with allowances for doubling in that year.) The resulting frequencies gave a total nonfarm income slightly less than that indicated by the statistics of national income paid out (.National Income in the United States, 1929-35, U. S. Department of Commerce, 1936), with allowances for agricultural income, for the income of individuals living in hotels, boarding houses, labor camps, and for net capital losses, dividends to insurance policy holders and similar allowances. “Family” incomes under $5,000, the distribution of which was based upon the Financial Survey of Urban Housing, were adjusted upwards by 5 percent in such a manner that both the number of households and the aggregate of all incomes were in agreement with the statistics mentioned above. The distribution of nonfarm households by income groups in the period 1935-36 was derived from Consumer Incomes in the United States, National Resources Committee. The basic data used in that report were obtained during the years 1935 and 1936 jointly by the Department of Agriculture and the Department of Labor. (The National Resources Committee report shows the income distribution of nonrelief families separately for farm and nonfarm families, but a similar break-down is not reported for relief families or for total families.) The distribution shown on the chart was derived by subtracting the number of farm “families” (relief and nonrelief), which are estimated to have been in each income class, from the total in each income class as shown in Consumer Incomes in the United States. Important differences with regard to the definition of income not fully covered in these notes exist between the distributions shown in figure 1. For example, in the distribution for 1935-36, income is defined from the point of view of expenditure and includes only the income received in a given year which is available for expenditure during that year. Differences also exist with regard to the definition of “families.” For example, in the above figure, in 1929 and 1933, single persons operating independent households were included, whereas in 1935-36 they were excluded. The 1929 nonfarm households include all families whose heads are not farmers; the 1933 nonfarm households represent all households other than those on farms. influence the family income available for other items of expenditure including shelter. The cost of shelter similarly influences the demand for other commodities. Summary of Demand Factors A graphical showing of two of the demand factors, the number of families added and average family income, appears in figure 2. With allowances for a lag of 1 to 2 years, the general correspondence between these measures of trends in the demand and the trends in the number of units upon which construction was started may be clearly observed. Some of the major exceptions in the correspondence between demand factors and building will be noted later at various points in the discussion of costs of ownership. A conspicuous exception occurred in the decline of building in 1920 in spite of an increasing number of families and increasing family income available for shelter. This development appears to have been largely influenced by the rapidly advancing costs of property ownership during 1919 arid 1920. The changes wrought by the World War disturbed in an unusual degree the adjustments which might otherwise have been expected. A detailed year-to-year discussion of the demand changes over these years is not practicable and can be more adequately stated in a quantitative formulation of major economic influences (pp. 10-12). It may be observed, however, that the sharp increase in families after the World War due to postponed marriages, farm-to-city migration, and substantial immigration from other countries, resulted in acute housing shortages in many cities in the United States. The number of families added in nonfarm areas which had been fairly steady at about 300,000 to 450,000 annually for the years from 1902 to 1917 suddenly dropped to slightly more than 250,000 families in 1918, and immediately after the World War advanced to approximately 450,000 in 1919 and to more than 600,000 in 1920 and 1923. The number of families added each year in nonfarm areas then declined almost without interruption from an increment of 650,000 in 1923 to 90,000 in 1932. Family income available for shelter (total income less allowances for food and other living costs) was well maintained over the period from 1923 to 1929. It declined sharply thereafter and did not show a substantial increase until 1936, the first year of any considerable volume of construction since 1930 and 1931. A substantial increase in the number of families added during the years 1934 and 1935 did tend to enlarge the demand during these years. However, just as in the immediate postwar years, it was not until family income and costs of ownership were favorable that substantial increases in construction occurred. As has already been stated, nearly all economic activity results 5 Housing Monograph made by the Construction and Real Property Section, Bureau of Foreign and Domestic Commerce. The data used in making the dwelling unit estimates for the years 1921 to 1937 are the building permit reports compiled by the United States Department of Labor. The income data are derived from National Bureau of Economic Research, Bulletin 66; America’s Capacity to Consume, Brookings Institution; Survey of Current Business, February 1938, Bureau of Foreign and Domestic Commerce; and from Dr. Wilford I. King’s estimates, Income in the United States, Its Amount and Distribution, 1909-19. Estimatag of number of dwelling units started in 1934, 1935, 1936, and 1937 were revised slightly, subsequent to preparation of chart. See table IV for correct figures. from multiple rather than from single causes, and the only satisfactory analysis that can be made is in terms of the aggregate effect of these influences. The Determination of Rent An aggregate measure of changes in the several elements in the demand, together with the corresponding measures of the supply of available units, may be demonstrated to possess a close relation to the trends in rents. An a priori analysis indicates that the trends in rents logically are a function of changes in the total number of families to be housed, the number of available units, family income and the competing claims of other items of consumption, particularly those of relatively inelastic demand. An analysis in quantitative terms of the behavior of economic society in the United States from 1913 to 1937 discloses a close correspondence between the trends in the underlying elements and trends in rents. It is realized that the data over this period for all of the variables mentioned above, particularly for years prior to 1920, are not entirely satisfactory and may be greatly improved in the future by more adequate reporting. Also, the actual determination. of rent is essentially a local problem. The second qualification, in view of the economic interdependence of the various parts of the United States and the resulting similarity in the trends in these measures throughout the entire country, is actually less important than it might first appear to be. The most widely used indexes of rents are those compiled by the National Industrial Conference Board and the United States Bureau of Labor Statistics. Both refer to rents paid by wage earners’ families. Resulting from the manner of its compilation, the National Industrial Conference Board index tends to be a measure of the rates at which new rental contracts are made, and the Bureau of Labor Statistics series an index of the changes in rents paid for dwellings of approximately the same facilities, age, and location from period to period. The former tends to lead the latter in the timing of its fluctuations by slightly less than a year. Both of these indexes with proper allowances for timing are believed to be closely related to the trends in average rentals actually paid for dwelling units of all types but do not directly measure such payments. The analysis described here attempts to measure, using the best data available for the nonfarm areas of the United States as a whole, the relationships which logically appear to exist between rent and the total number of families, the number of available units, family income, and the relative cost of competing items of expenditure. These relationships are finally expressed in an equation of the type: Xr=a 4*b Xo -J- cXj-1- d Xf 119118—39----2 6 National Resources Committee The first independent variable in this equation is expressed as an index of the ratio each year of the total number of families in nonfarm areas in the United States to the total number of available housing units in nonfarm areas in the United States. The second independent variable is expressed as an index of income per nonfarm family in current dollars each year. The third independent variable was not clearly revealed in the a priori analysis. After a careful study of the net influence of the several variables upon rent using graphical methods, it appeared that the cost of living other than for shelter was not significantly correlated with rent, but that the cost of food did evidence a negative correlation which is believed to be significant. Consequently, in the above expression, the third independent variable is an index of the ratio of the cost of food to the cost of the items in the cost of living, other than food and shelter. The degree to which rent appears to be explainable in quantitative terms as a function of logical related influences is quite close. This fact gives added force to the general observation made earlier in this section, that residential rents and construction activity are determined not by chance hut by measurable economic and other social influences. Graphical methods of analysis were used first in determining the net influence of each of the independent variables upon the dependent variable, rent. In each case the relation was so closely linear that an expression of simple linear type appeared appropriate. The constants in an expression of this type were calculated by algebraic methods3 yielding the following equation: Xb=-216.155+2.540Ao+0.778X-0.167Af. The data used in these calculations are shown in table II. The character of the residuals between the National Industrial Conference Board index of rents and the values derived from the estimating equation suggests that the latter might be improved slightly if a curvilinear relation were assumed between some of the dependent variables and the index of rents. It was discovered that a higher degree of covariation existed with a lead of 9 months between the independent variables and the dependent variable rent than that calculated for a simultaneous occurrence of these variables; that is, rent apparently is influenced by occupancy, income, and the cost of food 9 months prior to the period to which the rent measure applies. The most satisfactory period of lead, however, has not been fully explored in this analysis. The graphical methods indicate that a more satisfactory correlation could be obtained for a lead of one year for occupancy and 9 months for income. • See Mordecai Ezekiel, Methods of Correlation Analysis. Table II.—Data for calculation of indexes of rent [1921-1930=100] Year Xo—Index of ratio of families to available dwelling units1 Xi—Index of income per family1 Xr—Index of ratio of food costs to other items in cost of living • X«—Calculated index of rents* National Industrial Conference Board index of rents Residuals* 1914 96.84 60.25 120.46 56.5 59.0 +2.5 1916 96.98 58.87 120 04 55^9 59 0 4-3 1 1916 97.12 61.06 118.06 58.3 59.9 +1.6 1917 97.46 68.73 125 81 63 8 61 9 m1 9 1918 98.29 76.99 132 81 71 2 69 4 18 1919 98.75 85.79 123 30 RO 8 76 4 4.4 1920 99.53 94.60 116 60 90 7 91 2 1921 100.96 103.71 106.73 103.1 99.9 -3.2 1922 102.34 89 75 92.56 98 1 98 0 1923 102.88 94.16 94 63 102 6 102 2 1924 102.59 100.80 96.14 106 8 108 7 4-1 0 1925 101.45 100.23 97.53 103.2 106.4 +3.2 1926 100.13 102.46 104.45 100 4 103 5 +3 1 1927 98.88 103.34 106.33 97.6 100.0 +2.4 1928... 97.69 102.58 103.58 94.4 95.8 +1.4 1929 96.69 104.73 103.61 93.6 94.0 +.4 1930 96.33 105.42 103 89 93 1 91 5 1931 - 96.01 95.13 96.34 85.6 84 2 1 4 1932 95.92 79.01 82 86 75 1 74 0 1933 95.80 64.52 75.38 64.8 65.2 +.4 1934... 96.74 62.65 77 11 65 4 66 2 1935 98.61 68.97 83 64 74 0 71 9 T* o 2 1 1936 100.35 73.62 89.27 81 1 79 6 I 5 1937 101.57 81.36 90.72 89.9 88.4 -1.5 Preliminary extension of equation derived from data for period 1914-1937 gives the following figures for 1938 for the six columns above: 102.36,84.49,89.86,94.6,88.9, —6.6, respectively. i The ratio of total nonfarm families to the total dwelling units In nonfarm areas based on estimates made by the Bureau of Foreign and Domestic Commerce. Shown for annual period 9 months prior to that shown for rent. *The income per nonfarm family based on income data compiled by the National Bureau of Economic Research (Wilford I. King, 1909-18, and Simon Kuznets, 1919-28) and the National Income Section, U. S. Bureau of Foreign and Domestic Commerce, for the period 1929 to 1937. Includes adjustments for agricultural income and for direct relief payments. Shown for annual period 9 months prior to that shown for rent. Base period refers to series before application of lead. 8 The ratio of food costs to all items in the cost of living, excluding food and rent, based on the National Industrial Conference Board. Shown for annual period 9 months prior to that shown for rent. Base period refers to series before lead. ‘Based on the equation Xs=-216.155+2.540Xo+0.778X/-0.167X;r. • The difference between the actual rent indexes and the values derived from the estimating equation. Using the above equation, calculated values of the index of rents were determined. The actual and calculated indexes are shown in figure 3. The maximum discrepancy between the calculated and actual indexes is 4.4 points in the index of rent for the year 1919. Lesser discrepancies occur in other years. In view of the limitations which may exist in the data, the degree of covariation between the calculated and the actual indexes of rent is remarkably close. The substantial period of lead of the independent variables also may prove very useful in appraising future trends in rent in terms of related influences. The agreement between the actual index of rents shown in figure 3, and the values derived from the estimating equation fitted to these data, particularly in view of the assumption of linear covariation in simplest terms, is closer than might usually be expected in quantitative measures of this character and should not be taken as indicating the reliability of the particular formula shown above for pur Housing Monograph 7 poses of estimation.4 It does illustrate, however, the possibility of the scientific measurement of certain aspects of economic behavior with a degree of precision highly useful for many practical purposes.5 Economic Conditions Influencing Additions to Supply of Residential Units The economic process by which new units are added to the standing supply of available shelter is not essentially different in basic theory from that of other commodities. A number of important distinctions, however, must be observed with regard to residential building. The production of nondurable commodities on the supply side may be stated largely in terms of costs of production in the sense of costs of fabrication. The production of durable commodities such as residential building must be stated in much broader terms with the added consideration of such elements as interest and other financing charges, loss of value arising from obsolescence and depreciation, taxes, maintenance and other costs of ownership. A relatively small change in the standing supply of any durable commodity also may result in substantial expansion or contraction of credit and consequent large effects upon general purchasing power. Some of these points will be discussed more fully in subsequent paragraphs, particularly costs of ownership in their relation to building activity. Differences in Time Periods The customary distinctions of classical theory between current market, short-term, long-term, and secular trend conditions in the supply are also helpful in the analysis of fluctuations in residential building. This arises in part from the capacity of the basic demand to change substantially, in terms of the net annual increment in families, as well as to expand or contract greatly by doubling and undoubling, thus placing the requirements for new residential building from time to time under substantially different conditions. During certain limited periods, even substantial changes in costs of ownership do not appear to influence appreciably the volume of building activity. In other periods, as economic demand increases, cost of property ownership very soon plays an important part in determining the volume of construction of new units. In this case, short-term conditions, in which productive capacity is somewhat limited by available labor, materials, or investment funds, may in an important degree ‘ The coefficient of multiple correlation is 0.99 indicating that 98 percent of the variability (R*) over the period covered by this series is accounted for by the estimating equation. The significance of the regression coefficients is not precisely calculable for time series. Such tests as have been made, however, assuming each year to be a separate event, indicate that all of the regression coefficients including that of Xp are significantly different from zero, i. e., the likelihood of chance occurrence of a coefficient of the magnitude calculated is less than 1 case in 100. • In other fields, particularly agriculture, such methods have been used extensively. See Ezekiel, op. cit., pp. 337-340 for a list of correlation studies. influence the number of units which may be built. The immediate post-war period illustrates clearly the shortterm influences of cost upon the production of dwelling units. Such conditions were particularly marked in 1919 and 1920 and were repeated in 1937 and 1938. Long-term conditions in the supply permit certain adjustments in the number of workers and in the prices and production of necessary materials as well as in investment funds and business organization in the building industry. Secular trends, in building technology, in types and practices of lending institutions, and in living customs, e. g., from single-family to multiple-family units (see fig. 7), also result in important changes in production. The effects of such trends upon year-to-year production, however, are likely to be obscure and are not readily amenable to analysis in general terms. The quantitative measures of economic relationships developed in correlation functions shown later in this section reflect these changes in time periods, not as entirely separable experiences but as stages in a continuous process. Costs of Ownership The more important elements entering into the annual cost of property ownership are the cost of building and other improvements, the purchase price of land, interest rates and other financing charges, annual taxes and assessments upon land and improvements, and the annual loss of value due to obsolescence and depreciation. Supplementary costs, such as those for transportation and community services (refuse collection, recreational facilities, schools, and so forth) not adequately provided by public agencies, also enter 8 into any comparisons of the annual costs in different areas. Building Costs.—Improvements in the facilities included in houses have been substantial during the past two decades. Changes in residential building methods, however, have been very slight, and the technique of fabrication of houses has been one of the slowest of the arts to respond to the widespread technical progress of recent times. The result has been that a moderately rising long-time trend in the prices of building materials, accompanied by a much steeper trend in wage rates in the building trades, has resulted in a substantial increase in building costs during the past two or three decades. The failure of residential building to share in the technological developments which have made possible both high wages and lower costs in many types of manufacturing production has seriously hindered the improvement of housing conditions which would have followed substantial technical improvements and consequent reductions in the purchase price of dwelling units. Land Costs.—The costs of land and necessary improvements such as streets, water supply, and sewage disposal systems vary greatly in relation to the cost National Resources Committee of building. These site costs are influenced by many factors, such as transportation facilities, zoning regulations, and prospective uses of land other than for residential purposes. Financing Costs.—Interest rates and other financing charges, and the rate of retirement of invested capital made necessary by depreciation, obsolescence, and other loss of value, are also of major importance in any discussion of the production and supply of domestic shelter. Figure 4 provides a more direct comparison of the effect of interest rates and amortization upon various investments in housing which could be sustained by a given annual outlay for such purposes. This chart is a graphical representation of the mathematical relationship which exists between a capital expenditure, interest rates and periods of amortization, and the annual payment required to sustain that expenditure. It may be observed that equal annual payments of $240 for 20 years will sustain a capital expenditure of approximately $2,700 when interest rates are 6% percent, and that the same annual payments over 30 years with interest rates at 4% per cent will sustain a capital expenditure of approximately $3,900. Taxes, Figure 4. Source: Construction and Real Property Section, Division of Economic Research, Bureau of Foreign and Domestic Commerce. Housing Monograph 9 insurance, and other costs would, however, tend to be heavier under the larger capital expenditure.6 Taxes.—Taxes upon residential property, for the most part, are consumption taxes upon the users of domestic shelter. This is true not only for the 46 percent of the occupied nonfarm dwelling units in the United States occupied by the owners thereof, but also in all but a few cases, for rented properties, through the process of the shifting of the tax burden in the form of rental payments. Based upon a tax rate of $20 per thousand dollars of true value7 and an estimated value of residential property of 90 billions of dollars in 1937, the total amount of this tax paid in the United States may be estimated to be of the order of 2 billions of dollars annually. It may be alleged with some justification that these taxes, which are a large and important part of the revenue of local government units in the United States, are equitable charges for community services supplied by these government units. There are, however, notable exceptions to this observation. Although a tax may be levied solely for purposes of revenue, its broader effects are inescapable. The final incidence of taxes levied upon residential property may be stated in part as follows: (1) When a shortage of available units exists, new building is not likely to occur until the rental income from rented property (or the imputed income on owner-occupied property) rises to such a level as to exceed total costs of ownership including taxes. Under such conditions, taxes may be said to be shifted from the owner to the tenant. However, when the number of units available for use is considerably in excess of the current needs in a community, an increase in taxes cannot be entirely shifted by a property owner to the tenant. An increase in taxes upon real property under such conditions merely tends to reduce net income available from property, thus resulting in a reduction in an owner’s estimate of value based upon the capitalization of net income. (2) The incidence of taxes in a given area is also influenced by the differences in the tax rates in different parts of that area; for example, within and outside municipal boundaries. In immediately adjoining properties of equal desirability, it may be said that a tax placed upon one of the properties and not upon the other can not be readily shifted from the owner to the tenant. (3) In the event that taxation is universal and uniform upon residential property over a given area, the tax burden will tend to make property ownership more expensive. This will tend to reduce new building 6 For a further discussion of financing costs, see sec. V. 7 Facing the Tax Problem, Carl Shoup and Associates, 20th Century Fund, Inc., 1937. Dr. Shoup states that the property tax rate based on actual values “would probably exceed percent in the majority of cases.” The estimated true tax rate based upon the unweighted average of the rates in a number of cities over the period 1929-36 as calculated by C. E. Rightor and Rosina Mohaupt varied from $24.26 per thousand dollars in 154 cities in 1929 to $26.30 per thousand dollars in 230 cities in 1936, i bid., pp. 526-627. until, with increasing demand, the return to property is adequate to cover taxes as a part of the cost of ownership. Recognizing the effect of taxation of residential property upon home ownership, a number of States during the past few years have provided varying degrees of relaxation of the tax burden upon residential property, the most specific of which is designated as “homestead exemption.” Other things being equal, any reduction in costs of ownership such as taxes will make home ownership more attractive and under certain circumstances will tend to encourage new building. Other costs.—In addition to purchase price, financing charges, and taxes, a number of other costs such as those for repairs and other maintenance, and for insurance, supplementary expenditures for transportation and community services, as well as annual losses in value due to obsolescence and depreciation enter into the cost of home ownership. Purchase price, financing charges, and taxes, however, appear to be the dominant factors. Summary of Supply Factors The immediate postwar years illustrate clearly the effect of changes in building costs and financing charges upon residential construction. (See fig. 5.) During the latter part of 1919 and the early part of 1920, the costs of building materials moved up particularly rapidly, increasing 70 percent from April 1919 to the same month one year later, but declined equally abruptly, reaching a temporary low point in September 1921. Wages in the building trades advanced sharply in 1920 and increased somewhat during the following year but declined in 1922. Interest rates also advanced rapidly during 1920 and declined somewhat thereafter. In spite of an active demand during this period, a sudden rise in building costs and interest rates in 1919 and 1920 appears to have been largely responsible for the reaction in the volume of residential construction which occurred in 1920. In 1922, a year which experienced an astonishing increase in building activity, the aggregate of the major elements of cost was at a moderate level in comparison with immediately preceding and following years. The trends during the years 1936 and 1937 also illustrate the effect of increases in building costs upon building activity. Both skilled labor and common labor wage rates in the building trades which had been fairly steady during 1935 began to increase rather sharply in March 1936 and continued their steady advance to January 1938, after which they appear to have declined slightly.8 The prices of building materials as reported by the Bureau of Labor Statistics, after showing little change 8 Wage rates reported as actually paid by building contractors in 20 cities; cover common labor and 6 skilled building trades.—Engineering News-Record. 10 National Resources Committee The number of family units upon which construction was started annually was estimated by the Construction and Real Property Section, Division of Economic Research, Bureau of Foreign and Domestic Commerce. The basic data used in making these estimates for the years 1921 to 1937 are the building permit reports compiled by the United States Department of Labor. The index of building costs is based on selected items from railroad construction cost indexes published by the Interstate Commerce Commission from 1915 through 1937, extended to the years prior to 1915 using the trends in construction costs of wood, frame, and brick structures as reported by the American Appraisal Co. The index of long-term interest rates was based on the yields of 60 high-grade bonds as reported by Standard Statistics Co. Estimates of number of dwelling units started in 1934, 1935, 1936, and 1937 were revised slightly, subsequent to preparation of chart. See table IV for correct figure. for more than 2 years, also increased sharply in July 1936 and continued to advance until June of the following year. The substantial increase in building activity in 1936 and in the first few months of 1937 from the very low levels of preceding years was clearly arrested by the sharp increases in building costs and wage rates which began during the middle of 1936. On figure 5 are shown the fluctuations in interest rates and building costs, together with the number of dwelling units upon which construction was started over the period from 1900 to 1937. During periods of active demand the chart discloses that nearly all of the year-to-year fluctuations in building were in substantial agreement with theoretical expectation as regards the manner in which cost operates upon production. A more satisfactory composite measure of these fluctuations than is possible in the above terms is described in the following paragraphs. The Production of New Dwelling Units The economic conditions which appear to influence the volume of residential building have been outlined briefly in the preceding discussion. It is possible to carry this analysis farther and to express the relations between the volume of building and the related economic influences in terms of an estimating equation. The method of multiple correlation used in arriving at this equation was mentioned briefly in the preceding discussion relating to the determination of rent (see pages 5-7). The number of units built in a given period a priori appears to be influenced principally by the increments in the number of families in nonfarm areas, the average income of nonfarm families, the costs of ownership (particularly building costs, interest rates and taxes), and the condition of the market as measured by rents and the percentage of occupancy. The formula shown below is only one of the possible mathematical expressions for the relationships between these data but it expresses the relationship more satisfactorily than any of the others which have been developed as a part of this study. Log ^=-1.41624-.01352 XM+.OO428 A^+.O^OAc. The independent variables entering into this equation are: XM—a series measuring the conditions in the market for existing units derived by taking the product of the occupancy ratio and an index of rent changes for the annual period immediately preceding by 3 months that shown for the number of units upon which construction was started. XN—a measure of the number of families added during the year immediately preceding that shown for the number of units upon which construction was started. 11 Housing Monograph Xc—a measure of the ratio of income per family to costs of ownership for the annual period immediately preceding by 3 months that shown for the number of units upon which construction was started. In this ratio, cost of ownership is a consolidated figure derived by multiplying the sum of interest rates and taxes by an index of building costs. (See footnotes to table III.) In the analysis of rent changes (p. 6), a good fit was obtained using an estimating equation of linear type. The analysis of building activity suggested that a better expression could be secured using an equation of semi-logarithmic type as indicated above. This equation accounts for 93 percent of the variation (coefficient of multiple correlation squared) in the number of units upon which construction was started annually over the period from 1914 through 1937. In view of the complex nature of the variables and the extensive intercorrelation between the several variables, care should be used in evaluating the significance of this function.85 It is possible to translate changes in the independent variables into corresponding changes in the number of new units derived from the estimated equation. In view of the logarithmic character of this equation a change in the independent variables has a different effect at different levels in the magnitude of the variables. In terms of the 1921-30 average, a 10 percent increase in the index of conditions in the market (rent times occupancy) may be expected to be accompanied by a 37 percent increase in the number of new units started annually (the other independent variables assumed to be constant). Similarly, a 10 percent increase in the ratio of income per family to the costs of ownership may be expected to be accompanied by a 45 percent increase in the number of new units started, and an increase of 10 percent in the index of the number of families added, by a 10 percent increase in the number of new units. In figure 6 are shown the values derived from this estimating equation compared with the number of residential units actually built. As may be observed in this chart, the two series are in close agreement with the exception of the years 1918, 1919, 1930 and 1931. During 1918, a strong effort was made by the Government to discourage construction not directly required in the prosecution of the war. Consequently, the number of units started in 1918 is notably less than that which would have been expected had this restraint not existed. In 1919, with the relaxation of governmental war efforts, a reversal of these influences occurred and building volume was considerably larger than otherwise would have been expected. There is no clear reason for the discrepancy in 1930 and 1931. •• See footnote 4, p. 6. Table III.—Data for calculation of new family dwelling units started Year Xu Index of conditions in the market1 (1921-30 =100) Xir Number of families added > (1921-30 = 100) Xc Ratio of income per family to cost of ownership « (1921-30 = 100) Xu Calculated new units started annually4 (thou-(sands) New units started annually (thousands) * Residuals 6 (thousands) 1914 57.01 71.98 128.4 374 440 +66 1915 57.11 82.60 128.4 417 440 +23 1916 57.90 78.75 129.0 421 420 —1 1917 59.77 75.09 119.7 304 300 —4 1918 64.94 82.60 113.7 307 200 —107 1919 72.38 89.01 102.2 269 460 +191 1920 87.04 49.27 99.1 256 300 +44 1921... 100.21 82.60 91.2 399 432 +33 1922 100.47 115.75 92.8 591 676 +85 1923 102.21 105.86 97.9 684 814 +130 1924 109.32 109.52 96.9 853 827 —26 1925 108.13 119.05 99.8 1,006 894 —112 1926 103.79 97.44 103.3 809 841 +32 1927 99.68 100.001 102.5} 709 757 +48 1928 94.68 102.20I 103.9' 653 713 +60 1929 91.54 88.10 106.4 566 510 —56 1930 89.97 79.49 102.5 428 303 —125 1931 83.19 89.74 96.2 303 219 —84 1932 73.14 50.37 88.8 114 94 —20 1933 63.07 37.91 84.6 63 64 +1 1934 62.86 16.67 87.9 58 62 +4 1935 68.82 64.29 94.6 142 149 +7 1936 _ _ _ 76.83 94.32 101.9 321 300 —21 1937 85.93 83.33 98.3 335 327 -8 Preliminary extension of equation derived from data for period 1914-1937 gives the following figures for 1938 for the six columns above: 89.23, 83.33, 93.1, 303, 360, +57, respectively. i The product of the National Industrial Conference Board rent index and the occupancy ratio. The occupancy ratio prior to 1930 is based on the ratio of total nonfarm families to total available nonfarm dwelling units, adjusted so that ratio never exceeds 100. After 1930 the occupancy ratio is based on actual vacancy surveys conducted by real estate boards and other local organizations. Shown for the annual period 3 months prior to that shown for new units started. (See table II and fig. 9.) * The annual net increase in nonfarm families. Shown for the year immediately preceding that shown for new units started. * The ratio of income per nonfarm family to the cost of home ownership. The income per nonfarm family is based on income data compiled by the National Bureau of Economic Research (Wilford I. King 1909-18 and Simon Kuznets, 1919-28) apd the National Income Section, Division of Economic Research, U. S. Bureau of Foreign and Domestic Commerce for the period 1929 to 1937. Includes adjustments for agricultural income and for direct relief payments. The cost of home ownership is the product of the residential building cost index and the sum of the weighted average interest and tax rates. The residential building cost index is based on Interstate Commerce Commission building cost reports for construction projects similar to residences. The mortgage interest rate is that published by Mr. Roy Wenzlick in the heel Estate Analyst for May 1937. The tax rate is computed from data published in the Financial Statistics of State and Local Governments, published by the Bureau of the Census. Shown for the annual period 3 months prior to that shown for new units started. (See table II and fig. 5.) ‘ Based on the equation Log XCz=-1.4162+.01352X’m+.00428Xw+.01620Xc. 5 See footnote table IV. «The difference between the calculated and actual new units. Although a decline in income late in 1937 and during the first part of 1938 suggested a decline in the number of units added during 1938, the number actually added during this year was slightly larger than the number added in 1937. This is doubtless due in part to the substantial changes in provisions for insuring loans by the Federal Housing Administration upon a higher percentage of value and at lower interest rates than those prevailing in 1937. This analysis can be further improved by the compilation of more representative statistics for nearly all of the variables; the calculation of regression coefficients in terms of quarterly data, and the careful determination of the optimum lead for each of the independent variables. 12 National Resources Committee NUMBER OF NEW FAMILY DWELLING UNITS STARTED IN NON FARM AREAS Figure 6. Sources: See table III. Adjustment of Supply of Houses to Current Demand Although new residential building in all but the most inactive years is a large and highly important branch of industrial activity, it stands in relation to the total housing supply as a comparatively small increment in the number of available units. Even in the peak years of 1909 and 1925, as has already been indicated, the number of new units added only slightly exceeded 4 percent of the then existing supply, and in 1933 and 1934, it fell to less than 0.3 percent of the standing supply. Residential Building Activity in the United States Estimates of the number of nonfarm family units upon which construction was started annually in the United States over the period from 1900 through 1938 are shown in table IV. Although satisfactory reports are available as far back as 1900 for only a very few cities, by using considerable care in the weighting of these data it is believed that a fairly reliable measure of the number of dwelling units upon which construction was started has been secured. Based upon the estimates shown in table IV, and with allowances for demolitions and other losses and for conversions, the net increase in dwelling units over the period from 1900 to 1909 is calculated to have been 3,679,000 units. Over this same period using the Census statistics of families and the most likely figure for vacancies in 1900 and 1910, it would appear that the net increase in units should have been approximately 3,664,000. Similar figures for the period from 1910 to 1919,inclusive, were as follows: 4,014,000 units added by building and other changes in the supply and 3,993,000 units as indicated by Census statistics of families and assumed vacancies in 1910 and 1920; and for the period 1920 to 1929, inclusive, as follows: 6,494,000 units added by building and other changes, and 6,585,000 units as indicated by Census statistics of families and assumed vacancies in 1920 and 1930. Estimates of the dollar value of the annual expenditures for new residential construction over the period 1900-1937 and for the maintenance of residential structures over the period 1915 to 1937 are shown in table V. In comparing the magnitude of the expenditures over the entire period, it should be observed that building costs in 1915 and prior years were roughly half of similar costs during the 1920 to 1930 decade and roughly two-thirds of the costs during the years 1932 to 1936 inclusive. It should also be observed that the number of nonfarm families in 1900 was less than half of the corresponding number in 1930. 13 Housing Monograph Table IV.—Number of family units upon which construction was started annually in nonfarm areas in the United States, 1900-19371 [Thousands] Year Family units Year Family units 1900—— 240 1919. 460 1901 - — — 340 1920 300 1902 360 1921 432 1903 — 400 1922.... 676 1904 440 1923 814 1905 480 1924 827 1906 - 480 1925 894 1907 - 440 1926 . . 841 1908 440 1927—, 757 1909. 580 1928 713 1910... 500 1929.... 510 1911 — 480 1930 ... 303 1912 520 1931 219 1913 — 460 1932 94 1914 440 1933 64 1915 440 1934... 62 1916 420 1935.. . 149 1917 - 300 1936 300 1918 200 1937 327 Preliminary figure for 1938 is 360. Source: Construction and Real Property Section, Division of Economic Research, Bureau of Foreign and Domestic Commerce. 1 The estimates shown in table IV for the number of family units upon which construction was started annually in nonfarm areas in the United States from 1900 through 1914 were obtained from a compilation based upon published municipal reports for individual cities and upon the returns from questionnaires sent to a large number of cities. In this manner, data showing the dollar value of residential construction separately were obtained for 24 cities in the United States from 1900 to 1905; 31 cities over the period from 1905 to 1910; and 40 cities over the period 1910 to 1915. Using these data and making proper allowances for differences in geographic divisions and city size, an index of the dollar value of residential building was calculated for the period from 1900 to 1915, inclusive. This index was deflated for price changes using the American Appraisal Co. index of cost of wood-frame construction which was the type predominantly built over this period. Previous estimates of the number of dwelling units built in the United States for the year 1915 were extended to earlier years using the year-to-year changes indicated by the index arrived at in the above manner. The basic data used in making these estimates for the years 1921 to 1937 are the building permit reports compiled by the United States Deqartment of Labor. A discussion of the method used in arriving at these estimates for the period 1915 to 1937 appears in an article by the author on “Economic Factors Related to Residential Building,” The Annals of the American Academy of Political and Social Science March 1937, pp. 24-36. Similar estimates covering the period 1920-36 have also been compiled by David L. Wickens and Ray Foster, National Bureau of Economic Research Bulletin No. 65. For residential building alone, adequate statistics are not available for years prior to 1900. However, total building permits in a number of urban areas since 1830 are available and appear to have experienced cycles of large amplitude and with durations of 16 to 22 years similar to those shown in table IV for residential building.9 In figure 7 there are shown estimates of the percentages of dwelling units in one-, two-, and three-or-more-family structures upon which work was started annually in nonfarm areas corresponding to the totals in table IV. In view of the limited data available for the years prior to 1920, implicit reliance should not be placed upon the estimates for any particular year prior to that time. They are based, however, upon actual reports for several cities in each year and upon a detailed study of relationships between the characteristics in cities of ’John R. Riggleman, “Building Cycles in the United States, 1875-1932,” Journal of the American Statistical Association, June 1933, pp. 174-183. Further unpublished investigations which Mr. Riggleman has very kindly made available indicate that similar fluctuations also occurred in earlier years. Table V.—Expenditures for residential construction in nonfarm areas in the United States, 1900—1937 1 [Thousands of dollars] Year New construction Maintenance Total Private Public 1900... $350,000 (2) 1901 470,000 (’) 1902... 560,000 (2) 1903 620,000 (’) 1904. 700,000 (2) 1905. 880,000 (2) 1906 990,000 (2) 1907. 980,000 (2) 1908 920,000 (2) 1909 1,130,000 (2) 1910. 1,150,000 (2) 1911 1,010,000 (2) 1912 1,160,000 (2) 1913 1, no, ooo (2) 1914 1,010,000 (2) 1915 989,000 $240 000 $1 329 000 1916 . 1,108,000 350 000 1 458 000 1917 1 943,000 370 000 1 313 000 1918 717,000 $28,000 380,000 1,125,000 1919... 1,599,000 14,000 400,000 2,013,000 1920..... 1,609,000 420 000 2 999 000 1921 1,759,000 450 000 2 909 qoo 1922 2.832,600 480 000 3 312 600 1923 3,757,400 5J0 000 4 267 400 1924 4,300,100 560 000 4 860 100 1925 4,583,600 610 000 5 193 600 1926 4,590,600 660 000 5 250 600 1927. 4,288,600 710 000 4 998 600 1928 3,961,400 760 000 4 721 400 1929 3,423,700 820 000 4 243 700 1930 2,195,100 <840 000 4 035 100 1931 1,395,600 570 0OO 1 965 600 1932 641, 000 420 000 1 061 non 1933 313,900 370 000 683 900 1934 . 271,800 1 000 450 000 722 800 1935 533,000 9,000 580,000 1,122,000 1936.... 1,101,300 61,000 760,000 1,922,300 1937.. 1,393, 000 93,000 800,000 2,286,000 Preliminary figures for 1938 are: 1,285,000; 50,000; 800,000; 2,135,000. i For the years 1915-37 see, Construction Activity in the United States, 1915-37, Bureau of Foreign and Domestic Commerce, U. S. Government Printing Office, 1938. The difference between the number of family units upon which work was started and the expenditures for residential construction is explained fully in that bulletin. 2 Data not available. various size groups in the years for which more complete data are available. The chart is thus believed to disclose with sufficient reliability for all practical purposes the long time trends over the entire period. The percentage of one-family units shows a clear tendency to decline from nearly 80 percent in 1900 to less than 60 percent in 1916 followed by an interruption during the war and immediate postwar years and only slight changes from 1922 to 1928. Since the latter year, the increasing proportion of single-family units is conspicuous, indicating that the greatest rate of decline in building during the years of declining volume from 1928 to 1934 was in other than one-family structures. In 1936 and 1937, the family units in new one-family structures represented approximately 75 percent of all units built in nonfarm areas in those years. The percentage of units in two-family structures increased fairly steadily from 1900 to the war period and has 14 National Resources Committee declined fairly steadily since that time. The percentage of units in multi-family structures increased from 10 percent in 1900 to more than 30 percent in 1928; then declined to approximately 5 percent in 1932; and was approximately 20 percent during 1936 and 1937, years of moderate building revival. Conversion, Demolition, and Other Changes in the Supply Changes in the available supply of housing depend not only upon new building but also upon such factors as the conversion of large single-family houses to multi- RELATIVE IMPORTANCE OF TYPE OF DWELLING“ PERCENTAGE OF DWELLING UNITS IN ONE,’TWO,"AND THREE-OR-MORE'FAMILY STRUCTURES STARTED Figure 7. Source: Construction and Real Property Section, Division of Economic Research, Bureau of Foreign and Domestic Commerce. For the number of new dwelling units started annually to which these percentages apply, see table IV of this section. For the years 1920 to 1937, inclusive, the percentages of dwelling units in one-, two-, and three-or-more-family structures were estimated from the Bureau of Labor Statistics permit data covering most of the cities over 25,000 population and in recent years for an additional number of smaller cities. The cities for which reports are not available were assumed to have the same perceptage distribution as reporting cities of the same size. The percentage distribution in the small towns and villages (for which reports have not been made until quite recently) was based on the 1930 percentage distribution of occupied dwellings by types in small towns and villages, as reported by the Bureau of the Census. It was assumed that the several types of construction in the small towns and villages bore the same ratio to these types of construction in the smallest cities reporting permits as that which existed between occupied dwellings by types in these two size groups of cities in 1930. Prior to 1920 the number of reporting cities is small but believed to be fairly representative. Over this period the reporting cities were weighted according to size and geographic location, and adjustments were made for the fact that identical cities did not report every year. The percentage distributions by type of structure prior to 1920 were linked to the estimates for later years on the basis of a comparison with the years 1920 to 1923, inclusive. pie-family dwellings, the conversion of residential structures to other forms of use, principally to stores and offices, and the withdrawal of structures from use by demolition and by destruction through catastrophes such as fire, flood, and tornado. Statistics indicating the net change in the housing supply resulting from conversions are available for only a few cities.10 It is frequently assumed that the number of family units 10 The Bureau of Labor Statistics in its annual reports on Building Permits in the Principal Cities of the United ¡States published the number of families affected by alter- ations that changed family accommodations each year in the period 1921 to 1927, added by the conversion of large houses to apartments is approximately equal to the number taken out of use by the conversion of residential buildings to other purposes. Such information as is available, however, indicates that there was an appreciable net increase in units by this process over the period covered by these statistics, averaging possibly 50,000 units annually over the period 1920 to 1930 and some 30,000 units annually in 1934 and 1935. The demolition of residential structures in the United States in the past has been at a very low rate and over the period from 1920 to 1930 probably did not average more than 30,000 family units per year in all nonfarm areas.11 In the majority of cases during this period, demolitions occurred in connection with changes in land use from residential to commercial or other purposes. Only infrequently (prior to the last few years) have submarginal structures actually been taken down solely because they were no longer in demand. During the past four or five years, the rate of demolition has been substantially increased through the action of municipal authorities in prohibiting the occupancy of structures unsafe or otherwise unfit for use. In 1934 and 1935, this number averaged possibly 60,000 units annually, or twice the number taken out of use annually in the period from 1920 to 1930. The assistance of the Federal Government in the removal of such structures without cost to their owners has also accelerated the rate of demolitions. Withdrawal from use as the result of fire, flood, tornado, or other catastrophe may be estimated at approximately thirty thousand family units annually in nonfarm areas over the period from 1920 to 1929.12 Up to the present time in the United States, average I net changes by conversion, demolition, , and other with- I drawal from use have thus been small in comparison I with the number of new units built annually. During I the decade from 1920 to 1929, the number of units I inclusive. The number of cities included in these reports varied from 35 to 67. Conversion figures have also been published for Philadelphia by the Philadelphia Housing Association in its bulletin Housing in Philadelphia, 1932 for each year from 1923 to 1932 inclusive. The number of family units created by remodeling were determined in The Beal Property Inventory of the Cleveland ¡Metropolitan District by Howard Whipple Green for Cleveland and its environs from 1932 to 1936, inclusive. These three sources were the bases for the estimates of conversions. u Demolition rates estimated for the purpose of this chapter vary from 2.5 units per 10,000 population in 1926 and 1928 to 6.0units per 10,000 population in 1935. Frank J. Hallauer of the United States Forest Service uses a demolition rate of 5.0 units per annum per 10,000 based upon 1920 population for his computation covering the period 1920-30: “Population and Building Construction,” Journal of Land and Public Utility Economics, Feb. 1934, pp. 35-41, and Feb. 1936, pp. 12-18. Demolition data were obtained from the following sources: A Bureau of Labor Statistics survey in 1937 covering more than 100 cities over the period 1929 through 1935 and annual reports for 32 cities over the same period; the Philadelphia Housing Association in its bulletin Housing in Philadelphia, 1932 for each year, 1923 through 1932; the Annual Reports of Tenement House Department of the City of New York, covering multi-family units in New York City from 1918 through 1936; and the Beal Property Inventory of the Cleveland Metropolitan District, for 1932 through 1936. I2 Beport of the Committee on Statistics and Origin of Fires, National Board of Fire Underwriters, New York, May 28, 1936, p. 5; “American Homes Unnecessarily Sacrificed to Fire,” Fire Prevention Yearbook, 1925, p. 7. Housing Monograph 15 withdrawn from use by all causes probably did not exceed 10 percent of the number of new units built.13 A year-to-year comparison of the net increase in the number of families and the net increase in the supply of dwelling units resulting from building, demolitions, and other withdrawals from use and from conversions is given in figure 8. The corresponding number of total nonfarm families and total available nonfarm housing units at the beginning of each year are shown in table VI. This table indicates that adjustments in the number of units, as might be expected, occur rather slowly. Following the peaks in the number of families, the additions to the number of available units (largely due to new building) tend substantially to exceed the current needs. This phenomenon of over-building in periods of declining demand may be clearly observed in figure 8. Building reached a peak in 1910, 3 years after the largest increase in families in that period. The peak in actual building in 1926 also occurred 3 years after the peak in the annual increase in families in 1923. Similarly, after the turning point in the increments in demand in terms of families in 1918 and 1932, building failed to respond immediately. The ratio of families to available housing units continued to increase in the earlier period for 5 years and, more recently, » For a careful discussion of this subject see George Terborgh, “Present Situation of Inadequate Housing,” American Economic Review, March 1937, pp. 169-174. in 1938 was still increasing after an interval of 5 years. This delay in the adjustment of building to changes in the annual increments in families arises in part from excess vacancies which have to be worked off in such periods, and in part from the slowness with which property owners and builders fully realize that changed conditions exist. The latter was true in 1919 through 1921 and was equally true in 1938. One of the most important measures that can be taken to improve this adjustment of additions in the supply to meet changes in demand is the development of adequate statistical measures of vacancies in all cities of any considerable size in the United States. The development and prompt analysis of these vacancy statistics should prove extremely helpful in lessening both the shortages and over-building which appear to have occurred fairly regularly in the United States for more than one hundred years and appear to be characteristic of building in many other countries. Vacancy data and demographic statistics such as those for marriages, divorces, and migration are particularly inadequate in the United States at the present time. Vacancy statistics, although available in only a very limited number of cities, have been developed increasingly, particularly since 1930. Using only the cities for which vacancy surveys have been made in two or more consecutive years, there has been compiled the Figure 8. Source: Construction and Real Property Section, Division of Economic Research, Bureau of Foreign and Domestic Commerce. t Represents new units built plus units added by conversions minus units demolished minus units taken out of use by fire and other catastrophe. •Estimated average annual increase in nonfarm families, 1935-40. Estimates of net change in available dwelling units in 1934, 1935, 1936, and 1937 were revised slightly, subsequent to preparation of chart. See table VI for correct figures upon which calculations are based. 16 National Resources Committee Table VI.—Total families and total available housing units in nonfarm areas in the United States by years, 1900-1938 Year, Jan. 1 Total nonfarm families 1 Total available housing units Ratio of families to available units1 Thousands Thousands Percent 1900 10,025 10,285 97.47 1901 10,264 10,497 97.78 1902 10,519 10,758 97.78 1903... 10,849 11,077 97.94 1904 11,199 11,426 98.01 1905. 11,494 11,804 97.37 1906 11,855 12,231 96.93 1907 12,289 12,647 97.17 1908 12,761 13,074 97.61 1909 ... 13,052 13,480 96.82 1910 13,477 13,964 96.51 1911 13,840 14,478 95.59 1912 14,188 14,942 94.95 1913 14,581 15,415 94.59 1914 15,032 15,878 94.67 1915 15,462 16,301 94.85 1916- .. 15,872 16,714 94.96 1917 16,323 17,117 95.36 1918 16,809 17,451 96.32 1919 17,078 17,677 96.61 1920 17,529 17,978 97.50 1921 18,161 18,331 99.07 1922 18,739 18,673 100. 35 1923 19,337 19,212 100.65 1924 19,987 19,959 100.14 1925- 20,519 20,761 98.83 1926 21,065 21,593 97.55 1927 21,623 22,441 96.35 1928 22,104 23,222 95.19 1929 22,538 23,906 94.28 1930 23,028 24,472 94.10 1931 23,303 24,858 93.74 1932 23,510 25,078 93.75 1933- 23,601 25,213 93.61 1934. 23,952 25,248 94.87 J935 24,467 25,252 96.89 1936 24,922 25,302 98.50 1937 25,377 25,494 99.54 1938 25,832 25,779 100.21 Source: Construction and Real Property Section, Division of Economic Research, Bureau of Foreign and Domestic Commerce. * As indicated in the footnotes to table I, the term “families” differs from actual households. The latter can never exceed the number of available units. The number of families, as used here, however, in periods of acute shortage may result in doubling of two or more families into one household. The ratio of families to available units in such periods may be greater than unity. index of residential vacancies shown in figure 9. The percent of units vacant shown in this figure is somewhat lower than that disclosed in real property inventories in 1934, 1935, and 1936. The real property inventories in nearly all cases included all standing units designed for dwelling purposes regardless of whether they were then fit or available for occupancy. The vacancy surveys, made for the most part by real estate boards and chambers of commerce in various cities, frequently with the assistance of postal department mail carriers, cover in nearly all cases only the units which are habitable and available for use. These surveys are made in the several cities at different times throughout each year. Annual and semiannual averages fail to disclose significant interim changes, and the usual quarterly periods are quite unsatisfactory since their turning points at the end of March and Septem- RESIDENTIAL VACANCY IN NONFARM AREAS IN THE UNITED STATES Source: Construction and Real Property Section, Division of Economic Research, Bureau of Foreign and Domestic Commerce. ber come at the middle of the active spring and fall moving periods. Consequently, for the purposes of computing the data shown in figure 9, the reports have been classified into three groups, those made during January to April, May to August, and September to December, inclusive. Changes in vacancies result not only from differences in the relative number of units in relation to the changes in the number of “families” but also include and are markedly influenced by changes in the rate of doubling of two or more actual families into larger households. The increase in vacancies from 1930 to the latter part of 1932 may be attributed largely to the doubling o: families over this period and to some net migration back to farm areas during the year 1932. The decline in vacancies since 1932 has been due both to the undoubling of families and to a rate of increase in the actual number of families much greater than the number of units added. Prospective Future Increments in Demand Statistical measures of the physical needs for dwelling units, as well as forecasts of housing requirements a number of years in advance, have been made by some writers, based largely upon population and population increments with allowances for a changing number of persons per family.14 It is believed that conjecture based upon such measures and classification as marital status, sex, race nativity and age composition of the population projected into future years, will also yield useful results. 14 Warren S. Thompson, “Population Growth and Housing Demand,” The Annals of the American Academy of Political and Social Science, March 1937, pp. 131—137. Frank J. Hallauer, “Population and Building Construction,” Journal of Land ani Public Utility Economics, February 1934, pp. 36, 41, and February 1936, pp. 12-18. Housing Monograph 17 In table VII are shown estimates of the prospective number of families in the United States by 5-year periods from 1940 to 1960. For purposes of comparison, the table also gives the corresponding figures for earlier years based upon Census enumerations and the interim year estimates shown in table I. In view of the tendency of Census statistics by single-year age groups to cluster about 5 and 10 year ages, it was not believed desirable to calculate prospective increments in families by single years as was done for the years from 1900 to 1935 using other methods. (See table I.) Moreover, statistics of this character measure essentially long-term trend changes and have little validity in terms of singleyear expectations. The estimates of the prospective number of private families on January 1 of each 5-year period 1940 to 1960 are based upon estimates by Thompson and Whelpton of the future population of each race-nativity group by sex and by 5-year age groups. (Population Statistics, Pt. I, National Data, material prepared for Problems oj a Changing Population, National Resources Committee, October 1937, Government Printing Office.) The population estimates by Thompson and Whelpton have been made on several assumptions. For the purposes of this study, the estimates calculated assuming medium fertility and medium mortality were used. Recent studies by investigators in this field indicate that this basis of estimation may result in future populations somewhat larger than are likely, but that the excess of such estimates over more likely values is roughly equal to the prospective net immigration. Consequently, the slight net immigration which may be realized over this period has not been included in the calculations of the figures shown in table VII. In so doing, it is believed that the most likely future population estimates readily available have been secured. To each of these population estimates for 5-year age groups, by sex and by each of four race-nativity groups, there were applied 1930 age-specific marital status ratios to determine' the prospective number of persons in each marital status at each of the 5-year periods from 1940 to 1960. To the number of persons in each marital status in the several groups indicated above, there were applied head of family ratios to determine the number of families which might be expected in each period. The aggregate of the number of prospective families in the several groups gives the total number of families likely to exist at each 5-year period from 1940 to I960.15 The head of family ratios used were derived from the 1930 census. If the number of family groups consisting of more than one married couple living together tends to decrease in the future, the number of separate house 15 Unpublished estimates using somewhat similar methods have also been compiled by Frank Lorimer of the National Resources Committee and George Terborgh of the Division of Research and Statistics. Board of Governors of the Federal Reserve System. holds will tend to be higher than those shown in table VII. Also, if the percent married in future years tends to be higher than that in 1930, the number of prospective families will tend to be higher than that shown in the table. The converse results would naturally follow from opposite assumptions. Table VII.—Number of families in the United States by 5-year periods as of Jan. 1, 1900-1960 [Thousands] Year Nonfarm families Farm families Total families Average annual increase 5-year period beginning in year shown 1900 10,025 5,951 15,976 360 1905 11,494 6,280 17,774 462 1910 13,477 6,608 20,085 436 1915 . 15,462 6,804 22,266 40,5 1920 17,529 6,762 24,291 577 1925 20,519 6 656 27,175 519 1930 23,028 6,740 29,768 1 412 1935 8 24,467 »7.360 * 31,827 8 479 1940. 8 34,221 1474 1945 8 36,591 8 452 1950 8 38,850 8 405 1955 8 40,877 »354 1960 »42,649 Source: Construction and Real Property Section, Division of Economic Research, Bureau of Foreign and Domestic Commerce. 8 See footnote 3, table I 8 Estimated prospective figures. Estimates of the prospective number of families derived from actual measures of the present age composition of the population by fairly homogeneous groups, together with estimates of the probable expectation of fife and known data on percent married and head of family ratios in a given year, notwithstanding their uncertainties, are believed to result in measures of future increments highly useful for practical purposes of appraising prospective needs for shelter and other requirements of economic society. An estimate of the number of families on January 1, 1935, using the methods outlined above, indicated a prospective increment from 1930 to 1934, inclusive, averaging 425,000 annually. During the depression years, however, some marriages were delayed and a substantial doubling up of families occurred. The number of families actually added over this period as indicated in table VII averaged 412,000 annually or somewhat less than the prospective increment, very likely due to the delayed marriages during these years. The number of actual households in the United States, as determined by using available vacancy data, appears to have increased over these years at the average rate of 350,000 annually, reflecting the substantial increases in doubling of families in the depression years. Trends in the prospective increments in families shown in figure 10 and in table VII indicate a substantial 18 National Resources Committee annual increase for a decade or more in the physical need for dwelling units. In nonfarm areas, the corresponding increase of families may be slightly less than the figures shown for total families in table VII, depending upon the extent of the prospective internal migration.16 Based upon these data, it is very roughly estimated that there may be required annually an average of approximately 550,000 dwelling units in nonfarm areas and possibly 40,000 units in farm areas during the present 5-year period ending in 1944. This includes those necessary for the replacement of units taken out of use and allows for conversions in nonfarm areas. Figure 10. Source: Construction and Real Property Section, Bureau of Foreign and Domestic Commerce. See footnotes 1,2, and 3, table I. *Based on estimated prospective figures of the number of families by 5-year periods, 1940-60. fThe increase in total families is the sum of the increases in nonfarm and farm families. Therefore, when there is a decrease in farm families, the increase in nonfarm families exceeds the increase in total families. See footnote 2, table I. The prospective increments in families, although declining slightly, remain large during the following period, 1944-49. A decline in each subsequent quinquennium also is indicated, falling in 1955-59, to the lowest rate over the 60-year period for which these statistics have been compiled. In view of the fact that many other important factors must be considered, these data do not provide an adequate basis for a forecast but do measure one of the major factors influencing the trend in the production of new dwelling units. As far as actual building is concerned, it is likely that economic influences such as family income and costs of ownership may accentuate the prospective trends suggested by the changes from period to period in the number of families shown in table VII. Past experience also indicates that the peaks in building tend to lag a year or more after the peaks in demand. is According to the revised estimates of the U. S. Bureau of Agricultural Economics, release dated Oct. 27, 1936, the net migration from farms to cities, towns, and villages averaged 630,000 persons per annum over the period from 1920 through 1929. Begin- ning in 1930, it declined rapidly, but actual reversal of migration, i. e., a net movemen t to farm areas, occurred in only one year, 1932. During 1935, the net movement to cities, towns, and villages had again reached a very large number—386,000 persons. The above conjecture upon the future demand is thus stated solely in terms of increments in famflies without allowances for trends in family income and in competing items of expenditure. The number of famflies appears to be one of the most important elements of demand as indicated by the high coefficient of net regression between rents and the ratio of families to available units in the equation shown on page 6. Future estimates of famflies also appear to be the only element upon which quantitative measures for future years are possible with our present knowledge of social behavior. Methods by Which Government May Supplement Private Enterprise Public action in economic affairs is not only possible but at many points is an integral part of the social process. Economic institutions and practices are to a considerable degree defined in the legal pronouncements and administrative practices of government. These institutions and practices touch residential building at many points. Laws relating to the ownership of property and to the enforcement of mortgage and other obligations resting upon real property; the regulation of banking and other agencies for the extension of credit; the formation and enforcement of building codes in their influence upon technical methods and technological change ; the taxation of property and of income ; municipal regulation of existing structures to the end of maintaining public health and safety—all illustrate the far-reaching relationship between public action and residential building. Although it is true that private enterprise operates essentially within an economic structure to a considerable degree defined by government, every change in public action or effort by government to supplement private enterprise has its repercussions upon the general economic process. Consequently, unless private enterprise is to be completely supplanted, any action by either public or private agencies to be effective must give careful consideration to the manner in which economic society operates under present laws and social customs. In this section, it is not possible to discuss in any detail governmental action in relation to residential construction. There are merely mentioned here a number of the processes by which government or other collective action or individual philanthropy may supplement the usual processes of private enterprise: 1. Direct ownership of housing projects by government bodies or by other public corporations, Federal, State, or municipal. 2. Loans or grants by government to private or to other public agencies. Housing\Monograph 19 3. Regulation of limited profit associations. 4. Establishment or regulation of financing agencies. 5. Police power regulation: health, sanitation, safety, and building codes, zoning, industrial legislation relating to building trades workers. 6. Land development under direct government ownership. 7. Technical research in building materials and methods. 8. Technical design and inspection. 9. Promotion by government of cooperative relations between all elements of building process—design, finance, labor, production of materials, property owners, and others. 10. Remission of property taxes in whole or in part. 11. Provision of market information—vacancy data, rent indexes, financial and other business statistics, statistics on families and family income. Summary of Building Trends and Related Influences The primary purpose of this section has been to outline the economic and other social conditions which have influenced residential building in the United States during recent decades. As has been indicated at several points, the discussion is essentially an introduction to this broad problem. It contains a number of new statistical series and careful fisting of the arious influences and a preliminary formulation of some of the relationships in quantitative terms. In this discussion of economic and other social influences, the attempt has not been made to appraise housing needs from a humanitarian point of view, but rather to describe the manner in which our economic society operates. The study is thus essentially one of observing the economic behavior of the nonfarm population of the United States during the past four decades as regards the manner in which it provides new domestic shelter A more precise quantitative formulation of this process remains to be completed in a continuing study of this problem. An observation of the relationships which exist between such influences as increases in families, family income available for shelter, building costs, financing costs, and taxes, and the rate of construction of new housing units indicates that it is possible to formulate this relationship in quantitative terms sufficiently precise for many practical purposes. Insofar as building is determined by the free play of social forces in an economy not subject to material change, it is believed that some indication of future trends might be made from time to time, particularly if suitable statistics upon vacancies, rents, and other measures of the market for residential shelter, as well as statistics upon the distribution of family income and the migration of families could be made available. However, if institutional practices such as public policy with regard to the extension of credit, taxation of residential properties, or indirect Government subsidy to residential construction, change greatly, any conjecture with regard to the prospects of the future must be altered accordingly. There is some justification for assuming cyclical regularity in the long-time fluctuations in residential building. A number of studies indicate that the period of these cycles may be about 18 to 20 years in the United States. However, the causation of these cyclical fluctuations has never been clearly established, and two of them appear to be influenced greatly by violent fortuitous changes, namely, the Civil War and the World War. Although the regularity of these cycles has some reasonable likelihood, great reliance should not be attached to the precise period of their fluctuations. It is believed that the principal emphasis in any appraisal of prospective residential building should be given to the current trends in the economic and other social influences which possess a close, logical, and measurable relationship to this branch of construction. o