56 Operating Costs in Public Housing per month, the coefficient of .662 means that a dollar increase in routine costs is associated with a 66-cent increase in average rent. The increase is positive as expected, but it is well below one, implying that rent increases under the present system of rents tend not to cover cost increases fully. The other coefficients are positive as expected; but they are much smaller and much less significant than the coefficient of routine costs. The private rent coefficient of .097 implies that a dollar increase in "low-budget" private rents per month leads, on the average, to an increase of only 9.7 cents in public housing rents. The response of rents to the income and elderly variables is even smaller and again not very significant. Why does the income-rent relation turn out to be so weak? There are several possibilities. One is that the statistical shortcomings of the income variable discussed earlier bias the results in the direction of a small and insignificant coefficient. A second possibility is that the deductions and exemptions in many local housing authority rent systems and the existence of flat per-room charges in a number of others greatly weakens the income-rent relationship. The most likely possibility is that in the recent period of growing cost pressures, local authorities, concentrating on trying to adjust rents to cover costs, have relegated income to a more passive role. If tenant incomes rise significantly during a period of severe cost pressures, this argument holds that this rise triggers no special action by the local authorities. If tenant incomes do not rise at such times, however, then local authorities may feel compelled to revise rent schedules. The true relation then appears to be between rents and costs, not between rents and income. The pooled regression results of Table 14 reflect both differences between cities and differences over time. Table 15 separates these two kinds of change by presenting regressions based on 4year averages and regressions based on deviations from 4-year averages. The 4-year average regressions tend to measure long-run effects, and the deviations tend to measure short-run forces. The 4-year average relation to routine costs is higher than the Rent Relationships in Public Housing Table 15. DWELLING RENTS - 4-YEAR AVERAGES Variable Coefficients t-ratios 57 4 Year Avg. Deviations 4 Year Avg. Deviations pooled relation, but still below one. In the two relationships shown for deviations from 4-year averages (one including and one omitting the private rent variable), the coefficient for routine costs is far lower than in the pooled results. Evidently the relation of rents to costs is one that takes some time to adjust. The relation to private rents is about the same for city averages as for the pooled results, and not significant in either case. Private rents show up as a surprisingly important influence in the first regression, based on deviations from 4-year averages. Possibly a sharp rise in private rents makes public housing attractive to those with incomes above the public housing norm; if they then move into, or are not compelled to move out of, public housing, rental receipts in public housing may go up. The second regression based on deviations from 4-year averages is consistent with this substitution hypothesis about private and public housing. Since the hypothesis involves changes in income of public housing tenants as a link between higher private rents and increases in public housing rents, it can be expected (if the hypothesis is true) that omitting the private rent variable from the regression would cause the income variable to become more important and statistically significant. This is the outcome 58 Operating Costs in Public Housing of the final regression; it is the only one in which the income variable is a significant one. Even here, however, the coefficient of .033 implies a very weak response of rent to income. The substitution hypothesis is at best a minor qualification of the central finding that under the present public housing rent system as a whole, rents vary mainly with costs. Data Sources and A brief description of data sources and their limitations is helpful in judging the reliability of the study's results. The basic cost and rent data come from annual reports submitted to the Department of Housing and Urban Development (HUD) by each local housing authority [16]. Data on number of units under management come from the same source. These data follow uniform accounting procedures and are probably the most accurate in the study. There are, however, differences between cities in the time-span covered by each year's data, since different authorities report according to different fiscal years. The price and wage data came from three principal sources: Bureau of Labor Statistics consumer price indexes for the 23 cities [4], BLS studies of 1967 city-worker family budgets by city [5 and 6], and Census Bureau surveys of employment and payrolls of city workers, also by city [7]. The most important limitations of these data are probably conceptual ones; since they are collected for purposes other than studies of housing costs, they do not match the precise concepts or cover the exact items that would be 60 Operating Costs in Public Housing best suited to the present study. Thus, the consumer price indexes and their components (of which utility prices are used especially heavily in this study) indicate broadly which cities have low prices and which ones have high prices. But they do not refer to the precise items which housing authorities purchase in operating public housing. Similarly, the monthly earnings of employees in “common municipal functions” indicate broadly where city wage rates are low and where they are high. But they do not refer specifically to the wage rates of employees of local housing authorities. The data on tenant characteristics (including income) are based on re-examinations of tenants in public housing, conducted by each local authority and tabulated by HUD [15].' They are subject to significant margins of error for purposes of this study for several reasons. They do not cover tenants newly admitted to public housing during each year. Therefore, they do not reflect differences between the average characteristics of new tenants and other tenants. Elderly tenants are not necessarily re-examined during each year of occupancy, so there is some under-weighting of elderly characteristics and some variability in the degree of this underweighting from year to year. Finally, all tenant data for 1968 were based on only the first three quarters, since re-examinations conducted during the final quarter were not yet tabulated at the time the study was conducted. Because of these known shortcomings in the tenant characteristics data, reported values for a few characteristics in about half a dozen of the 92 observations2 were edited before the beginning of the statistical analysis. Items edited were those which met two conditions: (a) they were far out of line with tenant characteristics in the same city in adjacent years, and (b) the number of tenant re-examinations on which they were based was sig nificantly below the total number of units under management. Some unpublished data on tenant characteristics made available 1. These reports are not published, but some tabulations based on them appear in [11], [12], and [13]. 2. The 92 observations cover 23 cities for 4 years. |