2:2,,» 0' ABSTRACT RELATIONSHIPS BETWEEN SOCIO-ECONOMIC AND LOCATIONAL CHARACTERISTICS OF THE OCCUPANTS AND HOUSING CONDITION BY Dennis Udell Fisher The objective of this study was to examine the relationships between the socio-economic and locational characteristics of the occupants and housing condition on a national scale. Using data from the 1960 Census of Population and Housing, it was discovered that five characteristics had the largest effect on housing condition: (1) size of place, (2) occupational classification, (3) type of tenure, (4) education of the household head, and (5) household income. The sets of occupant characteristics which appeared to be most im- portant varied depending upon which measure of housing condition was used. However, these characteristics were usually the most signifi- cant. The magnitude and direction of these and other relationships are presented in the study. ~ The study includes estimates of both gross and net relation- ships, the net relationship is the effect of one characteristic with the effects of other characteristics removed. The effects of other characteristics are not removed from the gross relationships. Dennis Udell Fisher In pursuing the study objective an aggregate measure of housing condition was constructed. INDEX was formed by placing a value on, weighting, and summing the physical housing characteristics that are included in the Census. This measure is thought to be a more accurate national measure of housing condition than those presently used because: (1) it is more accurately determined, (2) it is more representative of general housing condition, and (3) it provides for more precise discrimination over a wider range of housing condition. During the construction of INDEX, the need to examine presently used measures of housing condition became apparent. It was deter- mined that the Census measure of structural condition and the mea- sure used by HUD, Standard and Substandard, are inadequate for most national policy decisions. They are gross measures, the one having three classifications and the other, two. They are inaccurate. And they may not represent general housing condition. The work done in this study indicate a need for a more adequate measure of housing condition and in some cases a re- direction of present housing policy. RELATIONSHIPS BETWEEN SOCIO-ECONOMIC AND LOCATIONAL CHARACTERISTICS OF THE OCCUPANTS AND HOUSING CONDITION B)’ Dennis Udell Fisher A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Agricultural Economics 1972 ,-,. ,3 , ACKNOWLEDGMENTS Many peOple have contributed to the consummation of this study. I wish to express my appreciation to my Guidance Committee: Dr. Dale E. Hathaway (Chairman), Dr. Lester V. Manderscheid, Dr. James T. Bonnen, Dr. Harry M. Trebing, and Dr. John P. Henderson for guidance during my graduate program. Special appreciation is also expressed to my Thesis Committee: Dr. Dale E. Hathaway (Chairman), Dr. Lester V. Manderscheid, and Dr. James T. Bonnen for guidance, criticism, and advise throughout this research project. Financial assistance for this study was provided through the Department of Agricultural Economics, Michigan State University. The data was provided through the Computer Institute for Social Science Research, Michigan State University. Thanks are also expressed to Carlton M. Edwards for advise during the study and Laura Robinson, Daniel C. Tsai, and Sylvia J. Samuels for computer programming. Special gratitude is expressed to my wife, Barbara, and children, Cheryl and Brian, for their patience and encouragement. Gratitude is also due the late Victor Nelson and his wife Elsa for moral and financial support. Any errors or omissions in this manuscript are solely the responsibility of the author. ii TABLE OF CONTENTS Chapter I. INTRODUCTION . Introduction The Problem Statement Objective of the Study . Resume of Previous Investigations Gross vs. Net Relationships. Measures of Housing Condition . Empirical Relationships . Procedure and Outline of the Study . The Data Used II. MEASURING HOUSING CONDITION . Introduction . Theoretical Considerations. A Theoretical Basis Appropriateness of the Housing Unit Public Policy vs. Private Demand . Characteristics Included. Criteria for Measurement. Problems of Measurement . Requirements of a Measure The Housing Condition Index Measures Included in the Index. Ranking the Measures in the Index. Weighting the Measures in the Index . The Index. Summary and Conclusions. iii Page 0‘1ka \lMU‘I 10 13 IS 15 16 17 18 19 20 25 29 30 32 33 37 39 4O 41 Chapter III. GROSS RELATIONSHIPS BETWEEN SOCIO-ECONOMIC AND LOCATIONAL CHARACTERISTICS AND HOUSING CONDITION . Introduction . . . . . Regions of the United States Empirical Results . Size of Place . Empirical Results . Location Within an Urbanized Area Empirical Results . Age of the Household Head . Empirical Results . Sex of Household Head Empirical Results . Race of Household Head . Empirical Results . Nativity and Parentage . Empirical Results . Metropolitan Residence in 1955 Empirical Results . Occupational Classifications . Empirical Results . Type of Tenure. Empirical Results . Education of Household Head Empirical Results . iv Page 42 42 45 47 50 52 58 58 6O 61 63 64 64 66 69 69 72 73 75 76 78 79 81 81 Chapter Household Income . Empirical Results . Dependency Ratio . Empirical Results . Summary and Conclusions. IV. NET RELATIONSHIPS BETWEEN SOCIO-ECONOMIC AND LOCATIONAL CHARACTERISTICS AND HOUSING CONDITION: PREDOMINANT INFLUENCES. Introduction . Model Specification . Abbreviated Models. Predetermined Variables . Endogenous Variables . Assumptions and Interpretation of the Models . Empirical Results. Y1: Six Rooms or More . . . . Y2:Structura11y Sound and Y3: Not Struc- turally Dilapidated. . Y4: Hot and Cold Water Piped Inside the Housing Unit . . Y5: Exclusive Access to Bath or Shower. Y6: Built from 1950 to 1960 Y7: One or More Bathrooms . Y8: Heating Equipment . Y9: Exclusive Access to Kitchen Facilities Ylo: Telephone Available Conclusions. . Net Relationship with INDEX Empirical Results . Weight Sensitivity of the INDEX . Procedure. Hypothesis The Models . Empirical Results . Conclusions . Summary and Conclusions. Page 83 84 88 88 90 92 92 94 94 96 106 108 109 109 114 121 126 128 133 138 142 146 150 152 153 157 157 157 158 162 167 168 Chapter Page V. NET RELATIONSHIPS BETWEEN SOCIO-ECONOMIC AND LOCATIONAL CHARACTERISTICS AND HOUSING CONDITION: NATURE OF THE RELATIONSHIPS . . . . . . . . . . 171 Introduction . . . . . . . . . . . . . . 171 Empirical Results. . . . . . . . . . . . . 172 Regions of the United States . . . . . . . . 173 Size of Place . . . . . . . . 177 Location Within an Urbanized Area. . . . . . . 180 Age of the Household Head . . . . . . . . . 181 Sex of the Household Head . . . . . . . . . 185 Race of the Household Head . . . . . . . . . 188 Nativity and Parentage . . . . . . . . . 191 Metropolitan Residence in 1955.. . . . . . . 194 Occupational Classification. . . . . . . . . 197 Type of Tenure . . . . . . . . . 201 Education of the Household Head . . . . . . . 203 Household Income . . . . . . . . . . . . 207 Dependency Ratio . . . . . . . . . . . . 210 Summary and Conclusions. . . . . . . . . . . 213 VI. SUMMARY AND CONCLUSIONS . . . . . . . . . . . 216 Measuring Housing Condition . . . 217 Relationships Between Socio- economic and Locational Characteristics of the Occupants and Housing Condition. . . . . . . . . . . . . . . 219 Opposite Patterns . . . . . . . . . . . . 220 Household Income . . . . . . . . . . . . 222 Education. . . . . . . . . . . 223 Occupational Classification. . . . . . . . . 224 Size of Place . . . . . . . . . . . . . 225 Type of Tenure . . . . . . . . . . . 227 Race of the Household Head . . . . . . . . . 227 Statistical Significance. . . . . . . . . . 228 Net versus Gross . . . . . . . . . . . . 229 Further Research Needs . . . . . . . . . . . 230 Measurement of Housing Condition . . . . . . . 230 BIBLIOGRAPHY . . . . . . . . . . . . . . . . . 232 vi APPENDICES Appendix I. II. III. IV. Representativeness of Structural Condition . Definition of Terms. Statistical Models . Accuracy of Measures of Housing Condition vii Page 236 258 270 278 Table II-l. II-2. II-3. 111-1. 111-2. III-3. III—4. III-S. III-6. III-7. III-8. LIST OF TABLES American Public Health Association, Dwelling Survey: Appraisal Items and Maximum Standard Penalty Scores . American Public Health Association, Environmental Survey: Appraisal Items and Maximum Standard Penalty Scores . . . . Measures of Condition Included in the INDEX of Housing Condition Measures of Housing Condition and Desirable Housing Characteristics Regions and Geographic Divisions of the United States . Percentage of Occupied Housing Units in Each Region of the U.S. With Selected Housing Characteristics. Percentage of Occupied Housing Units in Various Residence Categories With Selected Housing Characteristics . Percentage of Urban Area Occupied Housing Units Located in Central Cities or in the Remainder of Urbanized Areas With Selected Housing Characteristics . Percentage of Household Heads in Different Age Categories Whose Housing Units Have Selected Housing Characteristics Percentage of Occupied U.S. Housing Units With Selected Housing Characteristics by Sex of Household Head . . . . Percentage of Occupied U.S. Housing Units With Selected Housing Characteristics by Race of Household Head viii Page 21 22 34 44 46 48 53 59 62 65 67 Table III-9. III-10. III-11. III-12. 111-13. 111-14. III—15. IV-1. IV-2. IV-3. IV-4. IV-5. Percentage of Occupied U.S. Housing Units With Selected Housing Characteristics by Nativity and Parentage . Percentage of Occupied U.S. Housing Units With Selected Housing Characteristics by Metropolitan Residence in 1955. Percentage of Occupied U.S. Housing Units With Selected Housing Characteristics by Occupational Group of the Household Head Percentage of Occupied U.S. Housing Units With Selected Housing Characteristics by Type of Tenure . . . . . . . . . . . Percentage of Household Heads in Each Education Category With Selected Housing Characteristics . Percentage of Households by Type of Household and Specific Income Categories With Selected Housing Characteristics Percentage of Households in Specific Dependency Categories With Selected Housing Characteristics Binary Dependent Variables From Selected Measures of Housing Conditions Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and the Presence of Six Rooms or More . Estimated Net Relationships Between Socio—economic and Locational Characteristics of the Occupants and a Structurally Sound Housing Unit . Estimated Net Relationships Between Socio—economic and Locational Characteristics of the Occupants and a Housing Unit That is Not Structurally Dilapidated. Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and Hot and Cold Water Piped Inside the Housing Unit . . . . . . . . . . . . ix Page 70 74 77 80 82 85 89 107 111 115 117 123 Table IV-6. IV-8. IV-9. IV—lO. IV-ll. IV-12. IV-13. IV-14. V-2. V-3. Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and Exclusive Access to a Bath or Shower . Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and the Housing Unit Being Built From 1950 to 1960 . . . . . . Estimated Net Relationships Between Socio-economic and Locational Characteristics and One or More Bathrooms in the Housing Unit. Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and Four Desirable Types of Heating Equipment Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and Exclusive Access to Kitchen Facilities Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and Telephone Available. Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and INDEX Maximum and Minimum Multiple Regression Parameter Estimates from the Twenty Models With Changing Index Weights, Condition Index = f (Socio- economic and Locational Characteristics of the Occupants) Frequency Distribution of the Parameter Ranges From Table IV—13 by Their R/M Value Estimated Net Relationships Between Regions of the United States and Measures of Housing Condition. Estimated Net Relationships Between Size of Place and Location Within an Urbanized Area and Measures of Housing Condition. Estimated Net Relations Between Age of the House- hold Head and Measures of Housing Condition . Page 129 131 134 139 143 147 154 163 166 174 178 182 Table V-5. V-6. V-7. V-9. V-lO. V-11. V-12. A-I-l. A-I-2. A-I-3. A-I-4. A-I-S. A-I—6. A-i-7. Estimated Net Relationships Between Sex of the Household Head and Measures of Housing Condition . Estimated Net Relationships Between Race of the Household Head and Measures of Housing Condition . Estimated Net Relationships Between Nativity and Parentage of the Household Head and Measures of Housing Condition Estimated Net Relationships Between the Metropolitan Residence in 1955 and Measures of Housing Condition . Estimated Net Relationships Between Occupational Classifications and Measures of Housing Condition. Estimated Net Relationships Between Type of Tenure and Measures of Housing Conditions. Estimated Net Relationships Between the Education of the Household Head and Measures of Housing Condition . Estimated Net Relationships Between Household Income and Measures of Housing Condition . Estimated Net Relationships Between Dependency and Measures of Housing Condition Parameter Estimates from Canonical Correlation: Structural Condition = f (Other Measures of Housing Condition). Tests for Linear Relationships Between Structural Condition and Other Measures of Housing Condition. Structural Condition by Access to Kitchen Equipment . Structural Condition by the Number of Bathrooms . Structural Condition by Type of Water Supply . Structural Condition by the Year Built . Structural Condition by the Number of Rooms xi Page 186 189 192 195 198 202 205 208 211 240 245 251 251 253 253 255 Table Page A-I-8. Structural Condition by Type of Heating Equipment. . 255 A-IV—l. General Representation of Results of Original and Reinterview Surveys of Identical Persons . . . . 280 A-IV-Z. Measures of the Accuracy of Housing Characteristics . 284 xii CHAPTER I INTRODUCTION Introduction Food, shelter and clothing are often cited as some of man's basic needs. In an affluent society such as the United States, one would expect that these basic needs would be supplied. However, in 1967, 12.3 percent of all families in the United States had in- comes of $3,000 or less with 26.9 percent of all non-white families falling in this category [32, p. 198, Table 246]. Eight percent of the housing occupied by whites and 29 percent of that occupied by non-whites was considered by the United States Bureau of Census as structurally dilapidated or lacking some basic plumbing facilities [32, p. 272, Table 367]. These figures only suggest the well-known fact that some citizens in our society do not enjoy satisfaction of their basic needs. Fulfillment of these basic needs is important both to the individuals directly involved and to society as a whole. Adequate housing, in particular, can contribute to a man's sense of well being, productivity, income, and general health. The benefits go not just to the individual and his family but to the community as a whole. For example, with adequate housing rural and urban areas are more attractive to the eye, property values are higher, and citizens are less apt to be restless. Also, for many .\ 9" people there is a certain satisfaction in knowing that other families have adequate housing. Just as the benefits from adequate housing are broadly distributed so are the problems inherent with inadequate hous- ing. The individual and his family may experience discouragement, sick- ness and loss of income while the community appears blighted, restless, and the economic and social health of the area declines. Certainly housing is a vital part of man's relationship to his world. Public officials have exhibited a continuing interest in the quality of man's environment as is evident from their activities: zon- ing, parks, public utility systems, streets, city ordinances, welfare schemes, etc. The provision of adequate housing has been approached \/ 7_1 through building codes, FHA interest subsidies, rent supplements, slum clearance, urban renewal, and provision of low-rent housing as well as other plans. The President's Commission on Rural Poverty expressed concern over the condition of housing for the rural poor. ”They live in dilapidated, drafty, ramshakle houses that are cold and wet in winter, leaky and steaming hot in summer" [13, p. 93]. A number of federal agencies are vitally concerned with housing: Housing and Urban Development, Federal Housing Authority, the U.S. Public Health Service, the Farmers Home Administration, Housing Assistance Administration, the Bureau of Indian Affairs, Veterans Administration, and the Office of Economic Opportunity. This incomplete list of federal agencies could be supplemented by lists of state and local agencies and private groups. It is pre- sented only to illustrate a mounting concern which is calling for an extension of the American tenet of "equal opportunity for all" to both rural and urban housing. The Problem Statement In order to develop significant public policy in the hous- . . . . l 1ng area, research 15 needed to evaluate "object1ve1y" the extent of inadequate housing and delineate its determinants. The need for a measure of housing condition is emphasized by the Bureau of the Census: The development of reliable measures of housing quality has been one of the major concerns of the Bureau since housing statistics were first collected on a large scale in the 1940 Census of Housing. The concept "state of repairs" was used as an indicator of structural quality in the 1940 Census while the concept of "condition of structure" was used in the 1950 and 1960 Censuses of Housing [30, p. l]. Presently houses are classified in the Census of Housing as: (a) sound, (b) deteriorating (housing needing more repair than would be provided in the course of regular maintenance), or (c) dilapidated (housing that does not provide safe and adequate shelter and in its present condition endangers the health, safety, or well-being of the occupants). In this study we intend to construct an index of housing condition which includes an increased number of categories into which houses are placed, uses criteria that are more precise, and includes more dimensions of housing condition. Related to the need for a measure is the need to understand the socio-economic and locational determinants of housing condition. Understanding the causes of a phenomenon usually goes far toward suggesting means of altering it. However, this work is a statistical analysis of Census data and not a micro-level examination of 1Objective evaluation here means one in which the evaluator exercises as little personal judgment as the present state of social science allows. individual cases. Thus it will analyze relationships some of which are causal and some not. Those relationships between the socio- economic and locational characteristics of the occupants and dimensions of housing condition which are not causal still provide information for policy formation and evaluation. For example, the estimated relationships will help identify the characteristics of the target population. A housing program may be examined to see if, in fact, it is operated in such a way that participation by a portion of the target population is precluded. Objective of the Study The objective of this study is to examine the association between selected socio-economic and locational characteristics of the occupants and housing condition. This objective can be broken into two parts: (a) examine both gross1 and net2 relationships be- tween selected socio-economic and locational characteristics of the occupants and measures of housing condition that are included in the Census, and (b) examine the net relationships between selected socio- economic and locational characteristics of the occupants and a measure of housing condition to be constructed in this study (INDEX). 1Gross relationships refer to the relationships between two variables with the effects of other variables not removed. In this work these relationships are estimated using cross tabulations. 2Net relationships refer to the relationships between two variables with the effects of other variables removed. In this work these relationships are estimated using multiple regression and canonical correlation. Resume of Previous Investigations Previous studies have revealed several relationships be- tween socio-economic and locational characteristics of the occupants and measures of housing condition. In this section we present some of those relationships and characteristics of the studies reviewed. Gross vs. Net Relationships All of the studies reviewed, except Shurlock's [17], are based on cross tabulations of some socio-economic or locational characteristics of the occupants and measures of housing condition. Cross tabulations provide estimates of the gross relationships be- tween the variables being studied. The effects of other variables are not removed. Thus the estimated gross relationships usually represent the effects of the studied variables and some omitted variables. The objective of this study includes estimating the net effects of the socio-economic and locational characteristics of the occupants on levels of housing condition. These will be compared with gross relationships in order to examine their differences and similarities. Measures of Housing Condition The studies reviewed used a variety of measures of housing condition. The works employing Census data used mainly the Bureau of the Census classification of structural condition—~sound, deteriorat- ing or dilapidated, or a classification which can be derived from Census data by adding information on plumbing facilities--standard '- 1.. 9. g “a ”n or substandard. This latter classification system is defined as follows. A house is substandard if it is: l. Dilapidated, or 2. Lacks one or more of the following facilities: hot running water in the structure, flush toilet for private use, bathtub or shower for private use [30, p. 2]. The housing unit is classified standard if it is not substandard. Two reasons have been suggested for the wide use of this system. First, meaningful distinctions can be made on a nationwide scale. Second, the classification embodies the criteria of hazards to health, safety and welfare, the elimination of which has constituted the basic justification for legis- lation in this field [30, p. 2]. This latter system is more accurate than the Census measure of structural condition due to the addition of the plumbing facilities data which is more objectively determined information. Table III--7 reveals that the standard-substandard classification had a built-in correction feature. Of the estimated two million occupied units that should have been classified as dilapidated in the 1960 Census but were not, over one million were accurately reported as lacking plumbing facilities. Thus, the erroneous classifications of structural condition were in effect corrected by the plumbing facilities data [30, p. 19]. The Bureau of the Census classification system for struc- tural condition is known to be relatively unreliable. The inaccuracy of this measure is revealed in the Content Evaluation Study for Housing Characteristics (referred to as CBS) [25] reported by the Bureau of the Census. Only 33 percent of the houses classified as deteriorating and 38 percent of those classified as dilapidated in the CES reinterviews were similarly classified in the 1960 Census interviews. Many of the studies examined [13, 14, 15, 23] mentioned ‘the lack of objectivity and crudeness of this measure. The measure <3f housing condition constructed in this work is assumed to be more ‘1’. .o— v u .1. 5.. n. “I U" reliable because it contains the Census classification of structural condition, information on plumbing facilities, and other measures of housing condition included in the Census. An assumption about measures of housing condition appeared in many of the studies reviewed: different measures of housing con- dition are in fact highly positively correlated. Several single measures of housing condition such as structural condition and age of the structure were consequently viewed as representative measures for general housing condition. This assumption is examined in Chapter IV when the measure of housing condition, INDEX, constructed in Chapter II, is examined for weight sensitivity. The existence of this assumption is documented and a variation of it is examined in Appendix I--"Representativeness of Structural Condition." Empirical Relationships The studies reviewed provided empirical evidence regarding some of the relationships between the socio-economic and locational characteristics of the occupants and measures of housing condition which will be examined in this study. They are reported briefly below. I. Probably the most mentioned relationships when Census data are used are those involving regions of the country. Bird, Beverly, and Simmons using the Census measure of structural condition indicate that housing in the South tends to be less adequate than housing in the North [23, p. 4]. Pavlick and Coltrane note that: "Housing in the Appalachian Region is generally inferior to housing .- OD ... ... Un- - 5... h -. .- ~ OK‘D. ‘ “so- . "\ in the surrounding area and is below the U.S. average, according to the criteria on which this report is based" [20, p. V]. These studies and others [13, p. 93] indicate regional differences in housing condition. The measures used to represent housing condition were combinations of data on structural condition and plumbing facilities. If different measures had been used, the South may not have exhibited such a high percentage of housing in poor condition. 2. Some commonly used locational variables in any study of national housing condition are the residence categories--rural farm, rural nonfarm, and urban [5, 13]. It was noted in Rural Poverty in the U.S. that a higher percentage of rural housing is dilapidated than urban housing and that, in general, urban housing is more ade- quate for the old [14, pp. 44, 49]. Consistent with this conclusion, Bird, Beverly, and Simmons noted that 85.4 percent of all urban units were sound, while only 71.5 percent of all rural units were sound [23, p. 4] pointing to a general difference in housing condition be- tween rural and urban areas. A question which could be asked is whether in fact the same standards of housing adequacy are relevant for rural and urban housing. 3. The popular press has repeatedly indicated this third set of relationships: that racial discrimination results inpoorer housing conditions for non-whites. Hurst notes in a research publi— cation that non-whites are more likely to occupy substandard housing iJl South Carolina than whites, indicating that housing condition tends to differ depending upon the occupant's race [21, p. l] . Tl‘lese relationships were noted in several of the publications re- Viewed [13, p. 93; 5]. 21> 'be. n« ‘V— ti! £1» .‘h ‘ 9, e.“ In" .. dh~ '6 ”a 4. Income and housing condition were estimated to be positively but not linearly related [23, p. 5; 6, p. 55; 15, p. 12]. The indications were that the relationship was approximately linear to a certain level of income after which income showed little relationship to housing condition. 5. Bird, Beverly, and Simmons indicated that housing units occupied by owners had more bedrooms than those occupied by renters. They also noted that: "Owned housing was usually newer than rented housing" [23, p. 3]. A similar relationship was noted in The People Left Behind: "Rural families who rent are twice as likely to occupy substandard housing as families who own their homes" [13, p. 93]. The relationships between type of tenure and several measures of housing condition have been documented [5]. 6. Another set of relationships presented in The People Left Behind are between the age of the occupants and levels of housing condition. "A disproportionate number of the elderly occupy substandard housing in rural areas" [13, p. 93]. 7. Schaeffer and Edwards in a study designed only to construct a measure of housing condition suggest that there may also be a relationship between occupational groupings and housing condition [15, pp. 14, 15]. This hypothesis was not empirically tested. The relationships that have been listed above except rlumber 7 are supported by evidence from cross tabulations. In only Orle of the studies reviewed was an attempt made to determine net I‘Ealationship between the studied variables although this need was u.» . I‘. D" "n u 10 often cited. Using multiple regression techniques Hughes H. Spurlock examined the net relationship between prOperty value, years of edu- cation and income, and his measure of housing condition, "complete plumbing" or "incomplete plumbing” [17]. The work undertaken here will include estimates of both gross and net relationships between the socio—economic and locational characteristics of the occupants and measures of housing condition. Procedure and Outline of the Study, The primary objective of this study is to examine the relationships between characteristics of the occupants and levels of housing condition. We approach this objective by first developing an index of housing condition in Chapter II. This process involved choosing appropriate measures of housing condition from the Census data and combining them in a weighted index. The measure of struc— tural condition was to be used as a criterion for weighing the com- ponents of the index. This criterion was chosen for two reasons. No other criterion was found. Secondly, structural condition appeared to be generally accepted in the literature as a representa— tive measure for general housing condition.1 It was felt that using this procedure an index could be constructed which would be more accurate than the structural condition measure used alone and would allow for more levels of housing condition. The process of using structural condition as a criteria fTDr weighing the measures of housing condition to be included in the \ 1The general confidence in the representativeness of struc- Flliral condition, which is expressed in the literature, is documented 1r). Appendix I. ...¢ 1." l u .4. \.. . o O 5r. u; 11 index revealed empirical evidence that structural condition may not vary consistently with some of these other measures. As a result of this information and the general confidence in the representativeness of this measure expressed in the literature,1 structural condition was examined as a measure of general housing condition. This work is presented in Appendix I. Chapter III consists of a presentation of the estimated gross relationships between socio-economic and locational charac- teristics of the occupants and the individual measures of housing condition which are included in the aggregate measure, INDEX. Cross tabulations were used to estimate the relationships. Chi-square tests were used to test for the existence of a relationship. Chapter IV contains a presentation of the estimated net relationships between socio-economic and locational characteristics of the occupants and measures of housing condition. The measures of housing condition included in our aggregate measure are converted to binary variable. For example: Y1 = 1 if the unit has six or more rooms. 0 otherwise. Y2 = 1 if the unit is structurally sound. I 0 otherwise. Y3 = 1 if hot and cold water are piped inside. 0 otherwise. 1The general confidence in the representativeness of struc— tural condition, which is expressed in the literature, is documented in Appendix I. ..., In I:- [Y- \\. \P‘" cqu -v- 5... uSL' » 12 These binary variables are then used one at a time as endogenous variables in a multiple regression model with the socio-economic and locational characteristics of the occupants constituting the pre- determined variables. Another regression model involves the same set of predetermined variables with the index of housing condition as endogenous variable. A secondary objective of Chapter IV is to examine the INDEX for weight sensitivity. The weights on components of the INDEX are varied over a limited range while the INDEX is used in the regression model previously mentioned. The variations of the parameter esti- mates are examined for stability. This limited examination is not a conclusive test but does add some information relative to the question of weight sensitivity. Chapter V includes a comparison of net and gross relation- ships that are estimated and presented in Chapters III and IV. Chapter VI contains summary and conclusions regarding gross relation- ships, net relationships, and needed research. Appendix I includes an examination of the representative- ness of structural condition. Definitions of terms used in the Census data are presented in Appendix II. The statistical models used are described in Appendix III. In this section we have briefly covered the general pro- cedure and outline of this research. We now consider the data used to approach our objectives. 13 The Data Used The data used in this research come from the 1960 Censuses of Housing and Population. More specifically, The basic sample of the 1960 Census of POpulation and Housing was a 25 percent sample selected from the complete listing of all housing units and group quarters. For housing units and for persons living in housing units, the sampling unit was the housing unit and all its components. For persons living in group quarters, such as barracks and institutions, the sampling unit was the person [33, p. 20]. Specifically the data come from the one-in-a-thousand sample which: . makes available reels of magnetic tape or sets of punch cards containing the separate records of the characteristics of a 0.1 percent sample of the population of the United States as recorded in the 1960 Census. The names of the respondents and certain of the more detailed items on place of residence and some other characteristics are not revealed. Therefore, it has been determined that making records available in this form does not violate the provision for confidentiality in the law under which the census was conducted [36, p. 2]. The data were made available through the courtesy of the Computer Institute for Social Science Research located at Michigan State University. Further information relative to the data used is available through several Bureau of the Census publications [24, 33, 34, 35, 361- The accuracy of the Census data is discussed in several Content Evaluation Studies [25, 26, 27]. For purposes of this re- search, the data on socio-economic and locational characteristics of households are assumed to be reported without "content" error. Of the 179,563 persons included in the 0.1 percent sample tlno different groups of households and household heads are used as observation points in this study. This occurs because different F’Eirts of the Census observations are used. The total Census is made 14 up of the complete count, and a 25 percent sample which splits into the 5 percent and the 20 percent sample. The Census tapes used here, the 0.1 percent sample, are taken from the 25 percent sample and contain a 5 percent and 20 percent split. The parts of the Census used here are the 25 percent and 20 percent parts. Four thousand nine hundred thirty-four persons or 2.7 percent of the persons in the sample are omitted because they reside in group quarters. Also vacant housing units are omitted. Therefore this analysis is conducted using data representing 97.3 percent of the United States' population. Due to parity errors on the magnetic tapes as many as .4 percent of those observations used have been lost. We are assuming that this in no way biases the results. Using this data we will approach the problem of measuring housing condition in Chapter II and then move on to examining the characteristics associated with various levels of housing condition. CHAPTER II MEASURING HOUSING CONDITION Introduction Measuring the quality of things has captured man's imagi- nation for some time. We want to know the quality of our schools, our cities, our automobiles, etc. The more complex the thing being evaluated the more difficult quality assessment becomes. A metal part emerging from a machining process may be checked for dimension hardness and tensile strength. However, put a large number of parts together and the interrelationships between the various parts and the workings of the whole as well as the characteristics of the indi— vidual part become subject to evaluation. As one can imagine the complexity of the evaluation process increases rapidly as more pieces are added. The reader will note that housing is one of these things with sufficient component parts, the workings of which are confounded by the human element, that the multi—dimensional evalu- ation process is difficult. The primary objective of this chapter is to discuss the construction of an index which will more adequately measure housing condition on a national scale than the measures presently used. This chapter includes a discussion of the theoretical consideration 15 16 involved in measuring housing condition and a discussion of the construction of a housing condition index. Theoretical Considerations A literature review for a theoretical basis for measurement revealed several, some of which will be discussed later as they are relevant to this research. These bases varied in complexity, com- pleteness, and orientation depending upon the purpose for measure- ment. For example, the American Public Health Association (APHA) measure is designed to assess the healthfulness of housing over a city-wide area or part of a city. The Census of Housing measure, on the other hand, is designed for comparing levels of physical con- dition between areas of the country, race of the occupants, etc. In each case the dimensions of housing condition included, the relative weights and theoretical basis are different. In fact, there are as many theoretical bases as there are purposes for measuring housing condition. The measure to be constructed in this study is macro in orientation rather than micro as the APHA method. Because national Census of Housing data are used, the measure will be better suited to answering questions about the relative condition of housing between states or metropolitan areas rather than whether or not one area within a city should or should not be the subject of an urban renewal project. The constraint of national data from the Census of Housing effectively circumscribes the uses of this measure and in turn puts constraints on its theoretical basis. 17 A Theoretical Basis In choosing a theoretical basis, we examined the possi- bility of measuring housing condition relative to the specific occupant's well being. The very term customarily used to describe an assessment of housing condition "housing quality" implies some— thing about the well being of the occupants. This term is being purposely avoided because the measures examined and the one to be constructed in this work are an aggregation of specific housing conditions. In most cases there is no clear evidence that they reflect "housing quality" in general. One exception may be the APHA method which appears to reflect housing quality with respect to its standard, healthful housing. In this study the satisfying capacity of a housing unit is used as the basis for measuring housing condition. As the satisfy- ing capacity of a housing unit increases, the condition rating of that unit increases. This basis includes a wide range of the dimensions of housing conditions. Due to data limitations, this study is re— stricted to the physical characteristics of the housing unit. These characteristics are examined relative to their satisfying capacity for the occupants of the housing unit. If carried too far, this process leads to difficulty. For example, ceteris paribus, a ten-room house may have a higher housing con— dition rating than a four-room house for a family of eight people, while the opposite conclusion may hold if the housing unit is occupied by an elderly couple with no family. In the former case, 18 extra rooms has a positive effect on housing condition while in the latter they have a negative effect. Thus, household size and type could effect the condition rating of a housing unit for a specific household. Unfortunately, this type of paradox exists for many of the characteristics of housing units which would likely be included in an index of housing condition. We have not dealt with this paradox by rating each housing unit relative to the specific occu- pants. Instead each physical characteristic is rated according to how it relates to occupants in general. For "number of rooms," additional rooms are assumed to have a positive effect on housing condition. The assumptions regarding the affects of other physical characteristics are presented later in this chapter. Appropriateness of the Housing Unit Another question related to the satisfying capacity of a housing unit for its occupants is the question of general appropriate- ness of the housing unit. An example of this is furnished by housing units being located in different climatic and topographic regions of the country. The adequacy of any particular type of housing unit construction differs depending upon its location. National data does not assess these differences presently and accounting for them may not be feasible. The problem of appropriateness also occurs when the housing unit in question is not consistent with its surroundings. Examples of this are: (l) a mobile home in a residential district consisting of traditional housing units, or (2) a single family dwelling unit 19 among a group of multiple family units. In some cases, an incon- sistency in location affects the condition level of the housing unit in question. One final example of the problem of appropriateness occurs when examining the location of the housing unit with respect to the demand. Many homeowners are painfully aware of the financial loss associated with selling a house which is located in an area where demand has decreased relative to supply since the time of purchase. Two similar housing units can be sold for widely different prices, reflecting in part differing satisfying capacities of the two locations and resulting from differing supply and demand conditions. Five thousand housing units in eastern Montana would have a different money value than the same number and condition in New York City. Due to data limitations, no attempt is made to consider differences in housing demand at different locations. Public Policy vs. Private Demand Of course the purpose here is to assess housing condition from a public policy rather than a private demand point of view. If private demand were the basis, then our measure would focus only on those items which most affect the market price of the individual unit. As should be evident from the present interest in ecological problems, externalities can make private demand a poor basis for public policy. In reality, a national housing condition index will do little to measure the appropriateness of a housing unit with respect n \J 20 to the needs of a specific household, its location relative to the demand for housing, or its geographic or climatic setting. Instead the attempt is to measure housing condition with respect to its satisfying capacity for occupants in general from a public policy point of view and to include items in the measure that are not sensi- tive to climatic and geographic differences. Unfortunately, this involves omitting many items that are definitely related to housing condition. Characteristics Included A review of previous housing condition measures provides insight into the types of characteristics usually included. The American Public Health Association (APHA) method contains the largest number of characteristics. They are divided into two classes: (1) characteristics involving the housing unit itself which may adversely affect safety or essential livability of the unit, and (2) charac- teristics of the neighborhood. A list of these specific character- istics are presented in Tables II-l and II-2. Notice that the sc0pe and detail of the measure goes far beyond the data included in the Census. Also the items under "Occupancy" (Table II—l) show that a special emphasis is placed on the appropriateness of the housing unit for its present occupants. An examination of the Environmental Survey (Table II-2) reveals a substantial emphasis on the surrounding neighborhood and the appropriateness of the entire housing situation. In fact the measure includes so much information about the appropriate- ness of the unit that it might be better referred to as a measure 21 TABLE II-l.--American Public Health Association, Dwelling Survey: Appraisal Items and Maximum Standard Penalty Scores Item Maximum Score A. Facilities 1. Structure: Main Access 6 2. Water Supply (Source for Structure) 25 3. Sewer Connection 25 4. Daylight Obstruction 20 5. Stairs and Fire Escapes 30 6. Public Hall Lighting 18 7. Unit: Location in Structure 8 8. Kitchen Facilities 24 9. Toileta 45 10. Batha 20 11. Water Supply (Location and Type for Unit) 15 12. Washing Facilities 8 13. Dual Egress 3O 14. Electric Lighting 15 15. Central Heating 3 16. Rooms Lacking Installed Heater 20 17. Rooms Lacking Window 30 18. Rooms Lacking Closet 8 19. Rooms of Substandard Area 10 20. Combined Room Facilitiesb __;_-_ 360 8. Maintenance 21. Toilet Condition Index 12 22. Deterioration Indexc 50 23. Infestation Indexc 15 24. Sanitary Indexc 30 25. Basement Condition Index _JE§ 120 C. Occupancy 26. Room Crowding: Persons per Room 30 27. Room Crowding: Persons per Sleeping Room 25 28. Area Crowding: Sleeping Area per Person 30 29. Area Crowding: Nonsleeping Area per Person 25 30. Doubling of Basic Families 10 T27? Maximum Dwelling Total 600 aItem score is total of subscores for location, type, and sharing of toilet or bath facilities. bItem score is total of scores for items 16-19 inclusive. This duplicate score is not included in the total for a dwelling but is recorded for analysis. cItem score is total of subscores for structure and dwelling unit. \jSource: [1, p. 12] ,/ 22 TABLE II-2.--American Public Health Association, Environmental Survey: Appraisal Items and Maximum Standard Penalty Scores Item Maximum Score A. Land Crowding 1. Coverage by Structures 24 2. Residential Building Density 20 3. Population Density 10 4. Frontage Daylight Obstructiona _£1 0 B. Nonresidential Land Uses 5. Areal Incidence of Nonresidential Land Use 13 6. Linear Incidence of Nonresidential Land Use 13 7. Hazards and Nuisances from Industrial or Commercial Sources 30 8. Hazards to Morals and the Public Peace 10 9. Smoke Incidence _§ 2 C. Hazards and Nuisances from Transportation System 10. Street Traffic 20 11. Railroads and Switchyards 24 12. Airportsa 20 61? D. Hazards and Nuisances from Natural Causes 13. Surface Flooding 20 14. Swamps or Marshes 24 15. Topography 16 6'6 E. Inadequate Utilities and Sanitation l6. Sanitary Sewerage System 24 17. Public Water Supply 20 18 . Streets and Walks 10 54 F. Inadequate Basic Community Facilities 19. Elementary Public Schools 10 20. Public Playgrounds 8 21. Public Playfields 4' 22. Other Public Parks 8 23. Public Transportation-~Very Important in Rural Areas 6 24. Food Stores3 _;1 40 Maximum Environment Total 350 aProvisional item, not tested. Source: [1, p. 13] 23 of the appropriateness of housing for the area and the house- hold. Other measures of housing condition have also recognized the interrelationship between characteristics specific to the housing unit and those related to the environment of the housing unit. Schaeffer and Edwards divide up the characteristics included in their measure into environmental systems: (1) inside and outside structure, (2) plumbing, (3) electrical, (4) heating, (5) family activity, and (6) sight characteristics [15, p. 5]. This idea is stated again in a working paper prepared for the United States Bureau of the Census entitled, Measuring the Quality of Housing,] an Appraisal of Census Statistics and Methods [30] (referred to subsequently as Working Paper Number 25). We believe that indexes of housing quality can be readily constructed on the basis of objective data easily obtainable in a self-enumerative census of population and housing. The raw materials for the indexes are of two types: 1. Characteristics of the unit and of the structure in which the unit is located. These characteristics should have face validity. They should be readily recognized as housing characteristics. They would be such items as age of structure, lack of central heating, number of units in structure, availa- bility of plumbing facilities, availability of kitchen facili- ties, degree of crowding, etc. 2. Characteristics of the neighborhood in which the unit is located. These can be obtained as a composite of two kinds of data. The first is derived by assigning to each unit the average values for the neighborhood in which the unit is located (e.g., the block, enumeration district, or tract) of the charac— teristics that are obtained for each unit separately. Thus each unit would be classified not only as having all plumbing facilities or lacking one or more of them but as being in an area in which (x) percent or more of the units have all plumb— ing facilities. The second kind of data might come from direct observation of neighborhood attributes, although in the context of a decennial census there are considerable limitations on what can be feasibly done [30, p. 7]. 24 Most theoretical discussions of housing condition measures include both characteristics of the unit and of the neighborhood. The present Census measure includes only an assessment of the struc- tural condition (sound, deteriorating, or dilapidated) of a housing unit. A measure derived from Census data (standard or substandard) includes only a limited amount of information about plumbing facili- ties along with the structural condition information. Census data, to date, does not contain information regarding the setting of the unit and thus measures constructed from that data will be lacking those dimensions of housing condition involving the housing unit's environment. Also limited information on the number and type of con- veniences in a housing unit are documented in the Census. The data contains such information as hot and cold running water piped into the unit, just cold water piped in, water piped to the outside or no piped water. Also information is available on year built, toilet facilities, kitchen facilities, heating equipment, etc. Unfortu- nately, nothing is available regarding the quality of the specific facility and appropriateness to the unit in question. With the data now available, we will not be considering the quality of the original facilities, their present state of repair, aesthetic value or use- fulness. Instead the assumption is that in the aggregate assigning a value to physical facilities and including them in a weighted index will be a better measure of housing condition than the measures presently used. This assumption should be tested in further work. 25 Criteria for Measurement In order to measure housing condition one must have central criteria against which to compare different housing units. As is the case in this study, the theoretical basis usually suggest the apprOpriate criteria. However, two which have been used in the past deserve some further examination here. The first is that used by the Bureau of Census in assessing housing structural quality as sound, deteriorating, and dilapidated. The criterion used for this measure is the "health, safety and well-being" of the occupants of the unit. This type of a measure has its greatest discriminatory power when very low levels of housing condition are being considered. However, as interest is shifted to higher levels, one must consider criteria such as that level of housing which society deems desirable. The two criteria are not mutually exclusive but the former usually refers to a much lower level and may be a sub-part of the latter. The former refers to health and safety standards while the latter involves the social desirability of a particular level. The two criteria not only suggest different total levels of condition but also suggest the inclusion of different housing characteristics and different weights on those characteristics. The authors of Working Paper Number 25 stated that both of these criteria should be used. In broad terms, we have reached the following position with respect to standards of quality; there are two general standards. The first deals with the question, does the housing unit have any characteristics that are detrimental to health or safety? The second deals with the question, does the housing unit have any characteristics that do not meet minimum standards of well-being for its occupants? [30, p. 7]. 26 Also noted was that the first criterion mentioned--health, safety, and well—being of the housing occupants-~is more stable over time and a much more operational criterion for determining which characteristics will be included in the measure, their relative weights and the acceptable level for the aggregate measure. The second criterion relating to the social acceptability of various levels of housing quality is much more difficult to implement. While the first can be constructed by consulting a team of experts in the field of health, and safety as was done with the APHA method, the second relies on some kind of aggregation of the opinions of indi- viduals as to what is more or less desirable in housing condition [30, p. 3]. As one examines higher levels of housing condition, however, the ability of an aggregate measure based on the first criterion, health, safety, and well-being quickly loses its power of discrimi- nation. If we are to differentiate condition levels above the barest subsistence type housing, one must use a condition measure which is based on a criterion such as social acceptability. It seems to this researcher that the higher the level of housing con- dition at which one wants to discriminate the more difficult the task of obtaining a criterion to be used in selecting characteristics to include, weights to be administered, and levels of housing condition to be distinguished. The importance of having the "correct" criterion for weighting individual characteristics of housing condition seems to become less important when a large number of characteristics are 27 included in the index. This proposition is generally true because the larger the number of characteristics the smaller the weights on each individual one. However, this ignores the situation where a high proportion of the characteristics included may describe one particular dimension of housing condition to the neglect of others. Therefore, even in an index containing many characteristics attention should be given to the relative weights allotted to the various dimensions of housing condition. Annette Schaeffer and Carlton M. Edwards attacked the problem of finding a criterion for weighting individual character- istics by defining a number of needs which they felt a housing unit should provide for the occupants, changing these needs into environ- mental systems and weighting these systems equally. However, they give no justification for their weighting system. The American Public Health Association assess an individual unit with penalty points for deficiencies found in various charac- teristics of housing. The number of penalty points assessed, which is their weighting system, was determined by a group of experts, the Committee on the Hygiene of Housing [1, pp. 12-13]. Each reportable deficiency is graded according to the seriousness of that condition as a threat to health or safety or as a deterrent of comfort or general livability, in the judgment of these experts. -Thus, the score assigned to each appraisal item represents a consensus of experienced opinions as to the importance of that condition. These penalty points are usually adjusted to meet requirements of the area being surveyed. At this point the specific criterion and resulting 28 weights are set by the local authorities using the measure. The maximum penalty points recommended are presented in Tables II-l and II-2. This criterion and weighting system was not used in this research because the characteristics included in the Census data are not easily paired with characteristics included in the APHA method. In fact, a futile attempt was made to transfer the weights to Census data for comparison with the weighting system used in the measure constructed later in this chapter. It seemed that enough was lost in transfer to destroy any usefulness. Three other criteria, which have been suggested as a basis for choosing weights, will be mentioned here although they are not used in the research. These criteria weight various housing charac- teristics at: (l) the relative values that are used in assessor's manuals, (2) the importance used in condemnation proceedings, and (3) the relative new component prices. Both the first and the third criterion were not used for two reasons. First, both suggest weights that are subject to the quality of the characteristic itself and the Census data do not include this information. Secondly, both rely to varying degrees on a central criterion of private demand which has already been rejected for our measure. The second criterion, the importance used in condemnation proceedings, is felt to be too narrow for a general measure of housing conditions. We have chosen to weight equally the characteristics from Census data included in this study's measure of housing condition. Upon examining the data, our system seems as plausible as other 29 systems and not noticeably different from the APHA system. The specific weighting and procedures used to test it will be discussed later. Problems of Measurement Several problems make choosing relevant characteristics to measure housing condition difficult. First, as building techniques change, the patterns of defects in housing units change. Home im— provement operations which occurred in the decade prior to the 1960 Census serve to obscure many deficiencies which would have caused a housing unit to be listed as dilapidated. These include items such as wall paneling, aluminum siding, contact paper, and a host of other do-it-yourself home improvement materials [30, p. 14]. These changes make condition assessment more difficult. Finally, many characteristics may be poor measures of housing condition because they are included in safety, sanitary, and building codes. Some housing analysts believe that, because of the increased enforcement of housing codes since 1960, there has been wide- spread installation of inferior plumbing facilities in poor housing. This installation may be sufficient to classify low- quality housing as standard [30, p. 13]. The effect of these codes may be to bring into compliance those included characteristics to the neglect of other important characteristics. In this case a housing unit could be rated high based on code characteristics when possibly it should be rated low because of other defects. This would cause no problems if the codes contained all characteristics necessary to insure adequate housing. 30 However, they probably do not. The only safeguard is to include a cross section of characteristics not all of which are included in the various codes. Requirements of a Measure In order to be useful, a measure of housing condition should meet certain requirements. A list of five such requirements are presented in Working Paper Number 25 which was produced as part of an effort to improve the Census measure of housing condition. Since the measure constructed in this study used Census data, it is felt that these requirements are appropriate. 1. The statistics should reflect the "real" as opposed to the "apparent" state of affairs with respect to quality of hous- ing . . . thus the term "real" may be interpreted as ”rele- vant for the present circumstances and present uses." 2. The statistics should reflect real trends in the quality of housing. 3. The statistics should be comparable geographically. 4. The statistics should be built up from data obtained for individual housing units. 5. The statistics should be based on methods that distinguish various levels of quality of individual housing units [30, pp. 9-10]. It must be recognized that these five requirements would be difficult to test. Meeting these requirements depends upon the characteristics chosen to be included in the measure and the rela— tive importance placed upon each item. For purposes of this work, the information included in the Census is assumed to represent "real" housing condition and be comparable geographically. Of course, the data does come from individual housing units fulfilling requirement number 4. The index will be used to distinguish various levels of 31 condition and is assumed to be valid. The validity and nature of this measure will be discussed in more detail later. Another requirement, which is mentioned in Working Paper Number 25 is that the measure should have a minimum mean square error [30, p. 42]. This requirement is deceptively simple. Mini- mizing this statistic implies minimizing the sum of the following three items relating to the measure of housing condition: (1) the variance, (2) the square of the bias, (3) the sampling variance. Of course, in a sample as large as the Census, this third component is insignificant. The square of the bias, the second component, has to do with how well the items included in the measure reflect "real" housing condition, as well as the enumerator's ability to record various characteristics in an unbiased fashion. This component can be thought of as a measure of the accuracy with which the Census statistics describes the theoretical value of housing condition. The first component "the variance" has to do with the precision of the measure of housing condition; that is, the consistency with which condition is estimated. Needless to say meeting all of these requirements would be an impossible task to attempt here. The author will attempt in this study to make significant improvement upon the present census classification of sound, deteriorating, and dilapidated and the classification of standard and substandard while at the same time retaining the advantage of using Census data to construct the measure. This discussion of theoretical considerations has included such items as: (l) a theoretical basis, (2) appropriateness of the 32 housing unit, (3) public policy vs. private demand, (4) characteristics included, (5) criteria for measurement, (6) problems of measurement, and (7) requirements of a measure. Following this discussion we con— centrate on the objective of this chapter—-constructing an index which will more adequately measure housing condition on a national scale than the measures presently used. The Housing Condition Index The index constructed in this section is believed to be an improvement upon the Census measure of structural condition as a measure of general housing condition for several reasons. First, the index is believed to be more accurate. Bureau of the Census Working Paper Number 25 [30] indicates that the formation of the classification system, standard-substandard, by addition of plumbing information to structural condition markedly improved accuracy. This occurred because information that is more accurately determined was combined with structural condition. The index constructed here would contain the added information on plumbing facilities as well as other measures, all of which are more accurately determined than structural condition [25]. Second, the index is believed to be more representative of general housing condition. The theoretical discussion at the beginning of this chapter indicated that housing condition is a multidimensional concept with structural condition representing only a part. This index contains several dimensions of housing condition in addition to structural condition making it more representative than structural condition alone. 33 Measures Included in the Index1 The measures of housing condition that have been included in the index are presented along with the value assigned for each level of condition in Table II-3. This list represents only a part of the data regarding the housing unit, included in the Census. So few measures are available that choosing those to include and ex- clude becomes a process of using all measures that are in some way admissable. Some of those items omitted from the index are discussed here with brief definitions where necessary and reasons for their exclusion followed by an explanation of included measures. One of the variables excluded, persons per room [34, p. LVII], a crowding index, combines both household size and the number of rooms. This variable relates to the adequacy of a housing unit for a particular size household rather than to housing condition in general. The variable number of rooms was included instead. Characteristics relating to the value of the housing unit such as contract rent, gross rent, and value of property are excluded from the index for several reasons. First, these characteristics are not listed for certain housing units such as farms, nonfarm units with ten or more acres, or single dwelling units with an attached business [36, pp. 71-73]. Secondly, these characteristics respond to market conditions in such a manner that they would not necessarily vary consistently with housing conditions across several markets. 1The definitions of most of the measures included in the index may be found in Appendix II--Census Definitions. The Appendix con- tains excerpts from several Census publications and these original sources provide more detailed information [24, 33, 34, 35, 36]. 34 TABLE II-3.-—Measures of Condition Included in the INDEX of Housing Condition Number . . Valuea (j) Condition (Vj) 1 Structural Condition A. Sound 10.00 B. Deteriorating 6.67 C. Dilapidated 3.33 2 Telephone A. Telephone Available 2.00 B. No Telephone Available 1.00 3 Kitchen Facilities A. Direct Access, Exclusive Use 10.00 B. Direct Shared Access or No Equipment 6.67 C. Shared Access Through Another Unit 3.33 4 Water Supply A. Hot and Cold Water Piped Inside 10.00 B. Cold Water Piped Inside 7.50 C. Water Piped Outside 5.00 D. No Piped Water 2.50 5 Year Built A. 1959 through March 1960 9.90 B. 1955 through 1958 9.40 C. 1950 through 1954 8.50 D. 1940 through 1949 7.00 E. 1930 through 1939 5.00 F. 1929 or before 2.40 6- Heating Equipment A. Built-in Electric Units 10.02 B. Steam or Hot Water 8.35 C. Warm Air Furnace 8.35 D. Floor, Wall, or Pipeless Furnace 6.68 E. Other Means, with Flue 5.01 F. Other Means, No Flue 3.34 G. Not Heated 1.67 35 TABLE II-3.—-Continued . Number . . Valuea (j) Cond1t1on (Vj) 7 Number of Rooms A. Ten Rooms or More 10.00 B. Nine Rooms 9.00 C. Eight Rooms 8.00 D. Seven Rooms 7.00 E. Six Rooms 6.00 F. Five Rooms 5.00 G. Four Rooms 4.00 H. Three Rooms 3.00 1. Two Rooms 2.00 J. One Room 1.00 8 Bathing Equipment A. Exclusive Use of Bath or Shower 10.00 B. Shared Use of Bath or Shower 6.67 C. No Bath or Shower 3.33 9 Toilet Facilities A. Exclusive Use of a Flush Toilet 10.00 B. Shared Use of a Flush Toilet 6.67 C. Other or None 3.33 10 Number of Bathrooms A. Two or More 10.00 B. One and a Partial 7.50 C. One 5.00 D. Shared, Partial, or None 2.50 aValue is the amount assigned to a housing unit when it possesses one of the listed characteristics. Source: These measures of housing condition and the levels within each measure are defined in the Technical Documentation [36] and in Appendix II. The value assigned to the levels within each measure represent the author's judgment as to the im- portance of the levels. 36 A number of items included in the 5 percent sample were omitted because they were similar to information available in the larger 25 percent sample [36, p. 75]. For example, "number of bed- rooms" recorded in the 5 percent sample, is replaced in the index by a similar item "number of rooms" recorded in the 25 percent sample [36, p. 69]. Another item omitted from the index for this reason has to do with the heating system for the housing unit. "Fuel used for heating" [36, p. 75] from the 5 percent sample was omitted while "the type of heating equipment" [36, p. 71] from the 25 percent sample was included. Another group of items recorded in the 5 percent sample are omitted because they relate to facilities which are not permanently attached to the structure and are typically not left in the housing units when occupants change due to sale or rental. These items in- clude "clothes washers and dryers," "television and radio," and ”air conditioners and food freezers" [36, pp. 75-76]. The air con- ditioners can be permanently attached to the housing unit and thus not removed when the occupants move but their use is specific to certain areas of the country and in cold areas even the highest quality housing units may not contain such facilities. In summary, the reasons for rejecting the measures just discussed were: 1. The variable did not measure housing condition in general but specifically with respect to a certain type of house- hold. (pl .L T r.- .n.5 H; .‘ l -... ‘ ”3.5 ‘ ‘ I pp. \‘G‘ a . . .A’ F.- ‘b. 1H\ rilied A,“-‘ ‘“55€St ZeaSire 5933c n V ‘t “t: 'M H.‘ 37 2. The variable would not measure consistently housing condition across several markets. 3. The variables from the 5 percent sample measured almost the same condition as was being recorded by other variables in the larger 25 percent sample. 4. The variable records the presence or absence of facilities which are not usually permanently affixed to the unit and therefore should not be considered a part of housing condition in general. Those measures included in the index do not in general vio- late these four reasons given for exclusion. The possible exception is the inclusion of the data on availability of a telephone. However, this has been included because of the important part this facility plays in everyday life. The definition in Appendix II indicates that a telephone need not be inside the unit but must be available for incoming calls in order to be recorded as telephone available. Ranking the Measures in the Index The levels of housing condition within each measure are ranked ordinally as they appear in Table II-3 with ”A" being the highest level and progressing downward through the alphabet for each measure. 1. It is assumed for measure number 1 that a structurally sound housing unit is a higher level condition than a deteriorating unit which is of a higher level than a dilapidated unit. 38 2. For measure number 2 having a telephone available is assumed to be a higher level condition than no telephone available. 3. The third measure, kitchen facilities, has three dis- tinct levels for different types of access: (1) direct access, exclusive use; (2) direct-shared access or no cooking equipment; and (3) access through another unit. They have been ranked from best condition to worst condition as listed. It is assumed that direct-shared access or no cooking equipment is a higher level of condition than access through another unit. 4. The fourth measure, water supply, has four distinct levels of condition which listed from highest to lowest are, hot and cold water piped inside, cold piped inside, water piped out- side, and no piped water. 5. The next measure, year built, has six levels of con- dition with the newest units representing the highest level, the oldest representing the lowest level, and intermediate ages ranked accordingly. 6. The sixth measure, heating equipment, has seven discrete classifications with the bottom four being easily ranked. However, the top three categories: (I) built-in electric units, (2) steam or hot water, and (3) warm air furnace were not easily ranked. After consultation with Carlton M. Edwards, co-author of A Housing Quality MeasuringgScale [15], built-in electric units was ranked first and the next two were ranked equally. It may have been more correct to rank these tap three equally. The next level was floor, wall or pipeless furnace; then other means, with flue; followed by other means, no flue, and last not heated. ll} ‘ I ‘. ‘u .- I. H! L“ I“ 39 7. The next measure, number of rooms, is ranked with largest number of rooms being the highest condition down to the smallest number of rooms being the lowest condition level. 8. The highest condition level for bathing equipment was exclusive use followed by shared use with the lowest level being no bath or shower. 9. For the measure, toilet facilities, the highest level was exclusive use of a flush toilet, followed by shared use, and the lowest level was no flush toilet. 10. The last measure of housing condition, number of bath- rooms, has four discrete levels of condition. It is assumed that the more bathrooms, the higher the condition level. Weighting the Measures in the Index The next problem was choosing values to place on each level of condition. Some of the practical and theoretical problems associ- ated with selecting a system of values or weights have been discussed previously in this chapter. The resulting conclusion was, except for the availability of a telephone, to weight all measures in the index equally for lack of a better weighting system. The availability of a telephone has a maximum possible value of two if one is available and a minimum value of one if a telephone is not available, while the other measures in the index have maximum values of approximately ten. The telephone was weighted less because it was assumed to be less important. Notice that ignorance is assumed with respect to the relative values within each measure of condition. For example, 40 the four measures-~(1) kitchen facilities, (2) structural condition, (3) toilet facilities, and (4) bathing equipment--each have three levels of condition within them and the total possible of ten points is divided equally between these levels. The same practice is followed with respect to the other measures which have different numbers of levels in them. The Index The index is then formed by summing for each individual housing unit, the value received for each of the housing measures. where: i = the ith weighting system for the INDEX. j = the number of the condition measure as listed in Table II-3. V. = the value allowed for the jth condition measure as J listed in Table II-3. W. = the weight given to the jth condition measure. INDEX 21 is the one described here where all Wj = l, j = 1,2,...,10. The maximum and minimum possible scores for this index are 91.92 and 26.99, respectively. The actual maximums and minimums from our sample were 91.42 and 31.06, respectively. The mean score was 71.42 with a standard deviation of 10.26. INDEX 1 through INDEX 20 will be discussed in Chapter IV where they will be used to examine the INDEX for weight sensitivity. .. La» 4.. .L A 41 At that time more will be said with respect to the validity of the INDEX. Summary_and Conclusions The primary objective of this chapter was to discuss the construction of an index which will more adequately measure housing condition on a national scale than the measures presently available. In pursuing this objective, we have discussed a number of theoretical considerations which indicate the difficulties inherent in attempting to measure housing condition. Particularly troublesome are the problems of finding measures that are comparable between geographic and climatic areas as well as between rural and urban areas. The measures included in the Census need to be tested explicitly for comparability between these areas. We proceed then to develop an index of housing condition asserting that it is an improvement upon the present Census measure of structural condition as a measure of general housing condition for two reasons. First, it is more accurately and objectively determined. Secondly, it is more repre- sentative because it contains more of the dimensions of housing condition. Further discussion of this index is included in Chapter IV. Using a regression model which is developed there the index is examined for weight sensitivity and more can be said relative to its validity. In the next chapter, the gross relationships between the socio-economic and locational characteristics of the occupants and measures of housing condition are examined. The measures used are the ones introduced in this chapter and included in the index. CHAPTER III GROSS RELATIONSHIPS BETWEEN SOCIO-ECONOMIC AND LOCATIONAL CHARACTERISTICS AND HOUSING CONDITION Introduction The previous chapter dealt with the general topic of measur— ing housing condition. Selected theoretical matters associated with this illusive measure were considered and the measures of housing condition which are used in this research were discussed. Also the construction of the INDEX which is used in Chapter IV was discussed. In this present chapter we explain the socio-economic and locational characteristics of households that are used throughout the remainder of the work. The estimated gross relationships between these charac- teristics and the measures of housing condition discussed in Chapter II are also presented. In Chapter V these estimated relationships are compared to thelnet1 relationships which are presented in Chapter IV. 1Net relationships refer to the relationships between two variables with the effects of other variables removed. In this work these relationships are estimated using multiple regression and canonical correlation. 42 43 The primary focus of this chapter is to estimate and present the gross1 relationships between socio—economic and locational characteristics of the occupants and various measures of housing condition. Contingency tables are used to estimate the gross relation- ships between household characteristics and measures of housing condition. Each of these tables has been tested for the existence of a relationship between the variables but not for the direction of that relationship. The nature of the statistical test used does not provide information on the form, magnitude, or direction of the relationship. The null hypothesis being tested in each case is Ho: The probability of a housing unit having any level of housing con— dition is not affected by the characteristics of the household. See Appendix III for a further discussion of contingency tables and the statistical test being used. This chapter is divided into thirteen sections, one for each set of socio-economic and locational characteristics of the house- holds. Each section contains definitions of the household charac- teristics, hypotheses regarding their relationships to housing condition where necessary, and estimated gross relationships with various measures of housing condition. Each of the thirteen sets of socio-economic and locational characteristics were cross tabulated with nine different measures of housing condition. One summary table is presented for each of the 1Gross relationships refer to the relationships between two variables with the effects of other variables not removed. In this work these relationships are estimated using cross tabulations. III in .. h. 44 sets of household characteristics rather than nine cross tabulations. The summaries have been prepared by dividing each of the measures of housing condition at an arbitrary level. The percentage of housing units within a specific socio~economic or locational characteristic that possess the desirable housing characteristics or higher levels of housing condition are then reported. Table 111-1 presents the list of desirable housing characteristics that are used in the sum- mary tables and the measures from which they are derived. TABLE III-l.--Measures of Housing Condition and Desirable Housing Characteristics Measure of Housing Desirable Housing Condition Characteristic The Number of Rooms Six Rooms or More The Structural Condition Structurally Sound Water Supply Hot and Cold Water Piped Inside Access to a Flush Toilet Exclusive Access to a Flush Toilet Access to a Bath or Shower Exclusive Access to a Bath or Shower Year Built Built from 1950 to 1960 Number of Bathrooms One or More Bathrooms Type of Heating Equipment Heating Equipment Built-in Electric Steam or Hot Water Warm Air Furnace Floor, Wall, or Pipeless Furnace Access to Kitchen Facilities Exclusive Access to Kitchen Facilities ‘ Source: This table was constructed from data on the characteristics of housing included in the 1960 Census of Housing [36]. 45 Using this list of "desirable housing characteristics" we present definitions of household characteristics, hypotheses regard- ing their gross relationships to housing condition, and estimations of these gross relationships in the next thirteen sections. Regions of the United States The first characteristics presented here, regions of the United States, are almost always used in any national assessment of income, education, or housing conditions. As can be seen from Table 111-2, Northeast, North Central, South, and West the regional charac- teristics used, are such large aggregations of diverse areas that they are not adequate proxies for such things as climate, topography, or geography. However, it is felt that in the absence of better indi- cators, regions of the country should be used. Most studies which include this set of variables indicate that lower levels of housing condition exist in the South than in other regions of the United States [3, 13, 14, 22, 23]. Empirical results of these same studies indicate that income and educational levels are generally lower and that the population is composed of a higher proportion of rural residents and non-whites, all of which are thought to have a negative effect on housing condition. One of the questions that will be examined in this research is whether or not, after the effects of other characteristics have been removed, the net effect of the South on housing condition is negative. This will be accomplished through a comparison of the gross effects of regions of the United States with their net effects on levels of housing condition. No direct test of this question 46 TABLE III-2.--Regions and Geographic Divisions of the United States NORTHEAST REGION SOUTH REGION South Atlantic Division New England Division Maine Delaware New Hampshire Maryland Vermont District of Columbia Massachusetts Virginia Rhode Island West Virginia Connecticut North Carolina Middle Atlantic Division New York New Jersey Pennsylvania NORTH CENTRAL REGION East North Central Division South Carolina Georgia Florida East South Central Division Kentucky Tennessee Alabama Mississippi Arkansas Louisiana Oklahoma Texas WEST REGION Mountain Division Ohio Montana Indiana Idaho Illinois Wyoming Michigan Colorado Wisconsin New Mexico Arizona West North Central Division Utah Nevada Minnesota Iowa Pacific Division Missouri North Dakota Washington South Dakota Oregon Nebraska Califbrnia Kansas Alaska Hawaii 47 will be performed but a comparison of the empirical results should suggest an answer. Empirical Results These first characteristics of the occupants considered, region of the country, exhibited a significant relationship with each of the housing condition measures at less than the .005 level of significance. Two distinct patterns of relationships are observed (Table 111-3), the ”traditional" and the "opposite" relationships. The ”traditional" or expected one is where the West exhibits the highest percentage of housing units with the desired housing charac- teristics and the South the lowest with the North East and North Central regions being second third, respectively. This pattern of relationships appears with four of the desirable housing charac— teristics: hot and cold water piped inside, structurally sound, exclusive access to a bath or shower, and one or more bathrooms. Variations of this relationship appear with three of the other housing characteristics. The North East Region has the highest percentage of units with six or more rooms, followed by the North Central Region and Southern Region and the Western Region having the lowest percentage. The Western Region has the lowest percentage rather than the highest but the other regions follow the "traditional” pattern. A variation of the ”traditional" pattern also appears with the housing characteristic, exclusive access to a flush toilet. Here the relationship holds except for the Western Region having the second highest percentage of units with this condition rather than .HomH magmao: pew :ofiumHSQOQ mo mombmcmu coma .oHaEmm ucmoaom om .mmamu mHmEmm wcmmsocuum-:w-mco ecu :« oflnmfiflm>m meme memo: pflocemso: pcm mafia: wcfimso; :o mumw mo mcofiumfisnmu mmouu anm woumHSUHmo one: memmucmoaem ommch ”ouasom 48 m.oH m.m~ a.w~ e.mm mesogomso: .m.: mo «mascoopoa w.mm v.wm m.mm m.wm m.wm mmwufiafiomm cmnouflx ow mmmoo< m>flmsfiuxm H.mo m.mo m.ov m.mn o.ww oumcusm mmofiomfim no Ham: .poon newsman Hfl< Eye: hope: no: no Emoum ofihuoofim :Huuafism acmemwscm mcfiumo: m.mm e.mm H.4a o.ew N.Nm meooagumm who: go «so m.am m.a~ N.Nm a.m~ N.oN coma ou omaa scam sauna N.aw N.am 4.aa e.aw e.mm pmzocm Lo 55am m on amouo< m>bmssoxm m.ww 4.4m e.ma m.ww 8.4m eofisoe amuse a on amouo< m>bmssuxm H.mw a.om N.aa m.om 0.80 meamca eoaaa Laue: esou can so: a.Nw e.aw m.aa 4.4m n.8w venom aflsmwsuuaaum o.am N.mm «.mm m.am o.me use: no meoom xam Acofimem comm cw HmHOH mo uceopomv meaocomso: :onom :onmm cofimmm :oflmom ommm.: mo pneumoz campusom Hewecmu ummm mofiumwnmuumpwcu mcwmso: oHnmafimoo ucmuama :uuoz cuuoz moflumwnouomhmnu mcfimsor weuumfiom new: .m.: on» mo coflmom comm cw mafia: mcwmzo: beamsuoo mo mmma:oouom--.m-HHH m4mm meme memo: ufiocomso: can mean: wcfimao; co «one we chMHmfianmu mmouu scam woumfisoflmu one: mowmucoeuma omech neuusom o.- 5.0 m.o m.o 0.5 m.m n.m m.m v.v v.m o.- n.o mvflozemso: .m.: we ammucoouea v.5m n.5m w.nm n.wm v.wm w.mm ”.mm o.mm v.mm o.mm m.mm m.mm mofluflfiwumu conuufix o» mmouo< o>wm3~oxm o.mm v.~n w.~n m.~n n.wn m.mn v.mn m.mo m.mo 0.0m o.mw m.nm mumcusu mmofioafla no ~Hm3 .uoo~m oomchzu ufi< sum: noun: uo: no sebum ewauoedm cw-ufiwsm acosnfinvm mcwumo: v.mm m.oa 0.0m m.om N.mm N.mm m.Hm m.mw o.nw m.mm H.Hn m.mm mEOOhcumm one: no 0:0 H.vH v.mH m.mm v.v~ m.mm n.~m m.Nm m.o~ N.wN m.mm m.~m o.mH com" on Ommfi Souk uywaa m.vm m.~m m.Hm N.mm v.vm o.vm m.Nm m.~a m.mm m.om w.mn o.no Hologm no spam a o» mmouo< e>wm=~uxm n.vm ~.vm m.mm o.mm n.mm v.0m w.vm n.mm m.~m n.0m m.mn h.No HOHMOH :m=~m m o» mmoeu< o>wm=~oxm n.wm m.mm m.mm m.vm v.0m m.mm N.vm v.Nm m.mm n.0m o.vn m.oo ovfimcu comma noun: vHou can uo: H.5w v.mw N.vw v.vw m.mm o.wm m.om m.mm v.mw o.~m n.vn m.mo vczom >-mh3uu§hum o.o~ m.om N.mN N.mn w.mn n.5m 0.0m c.am w.mm v.~v N.nm 5.5m 0&0: ho maoom me Axuomouau comm :« amuOH mo acouuonv moon”; .coo.c¢o.a mma.mam oma.mmv mam.mv~ mam.ma mmm.a4 mam.v~ ama.o mma.e mo oeamuso auamcoz saga monumauoauauagu -ooo.oom -ooo.om~ -ooo.ooH -ooo.om -ooo.m~ -ooo.o~ -ooo.m -oom.~ aneuauuop Hausa sense «apnea: canaunmoo cash: newunmuouuauazu «cameo: wouuodom and: mowuououau oucovwmom unawua> cw mafia: ucwmso: vowmsouo mo oun»:oupom--.v-~mu mqmm moms memo: anon -omso: ocm mans: mcnmso; co mnmw mo mconnmflsnmn mmono Eonm oonmnzonmo one: mommucoonom omogh ”monoom 59 o.mo 0.5m monocomso: mon< coon: mo ommncoonoa «.mm w.nm monnnnnomm cogonnx on mmooo< o>nmsfioxm H.0w o.nn oomcnsm mmonomnd no Ham: .noonw oomcnsm nn< Enmz none: no: no Emonm onnnoonm annnnnsm ucoemnsom mannmo: m.om N.Hm mEoonsnmm onoz no one H.me w.mn coma on ommn Eonm unnam H.5m w.mm nozocm no :umm m on mmooo< o>nmsnuxm 0.5m m.em nonnon cmsam m on mmooo< o>nmsnoxm o.nm N.om oonmcH oomnm none: onou one no: m.mm m.vw venom xHHmnanoznnm o.Hv n.mm onoz no mEoom xnm flxnomonmu zoom on Honon mo ncoonodv nano npno Hanncou Hmnncou monnmnnonomnmgu wcnmso: onnmnnmoo an 802 an monnmnnonomnmcu mcnmso: oopoonom nun: moon< pouncmnn: mo noocnmEom man an no monnnu Hmnncoo an nonmooq mung: mcnmso: ponmsooo mon< :mnn: mo ommncoonomuu.m-HHH mgmm oboe mono: onozomso: can mane: mcnmsoc co «you mo m:0numnanmn mmonom Eonm nonmnoonmu one: moumncoonon omonh “oonzom o.N m.m m.m m.o m.5 5.w m.m 5.oH n.nn m.nn n.on n.m 5.4 q. monozomzo: .m.= mo owmucounoa n.5m N.wm m.mm o.wm o.wm w.wm 5.mm “.mm o.am m.mm n.mm o.wm w.5m n.mo monnnnnoma cozonnm on mmouo< o>nm=nuxm ~.om 5.mm o.~o n.vo m.mo n.5o o.O5 o.o5 o.n5 H.M5 m.~5 5.mo w.oo o.mm oomcnsu mmonoanm no nnm3 .noonm momenou n2 Em: noun: no: no smonm ennuoonm en-nnn:m ncoeansom Manama: o.m5 0.05 o.nw 5.nm o.mw o.vm 0.0m o.om m.5m o.mw o.mw v.5m 5.mw n.45 meooncumm onoz no «:0 N.o m.m «.mn o.vn ".mn n.m~ o.- m.m~ o.vm 5.ov o.ev o.oe “.mm n.0m com" on own" song unnsm n.55 m.o5 N.mm ~.vw o.mm ~.om m.5w m.5w ~.mw 0.0m 0.0m m.mm N.mm 0.55 noxozm no span a 0» mmooo< o>nmonoxm n.ow o.~w m.mw o.ow n.5m v.5m m.mw m.wm m.om m.nm o.nm ~.om v.0m o.ow nonnoh Among a on mmoou< o>nmsnoxm m.om n.nw «.mm o.cm o.mm 5.5m m.mw 0.0m m.om m.no m.nm n.nm “.mw c.5w common wound noun: vnou one no: n.¢5 v.m5 5.m5 o.ow 5.~m 5.mm n.vw m.mw m.vm w.mw o.vm «.mm v.om m.M5 venom nnnmnanoanum n.nv o.on m.om o.mm m.5m H.wn o.ov o.~v 5.mv ~.~v m.nm o.nN o.w n.m onoz no maooa xnm annouonmu om< comm an Hanan mo ucoonoav m.mm m.m5 m.V5 o.mo m.vo m.mm m.vm m.mv m.ve m.mm m.vm m.m~ m.vm m.mn -om -m5 -95 -mc uoo -mm -om -mv -ov -mm -om -mN -o~ -mn one: vnozomso: mo ou< monumnnouoanocu mannaoz unpannmoo monummnonuanugu ucnnsoz wouoonom o>a: mafia: ucnmooz omen: monnomouau ou< uconomunn an memo: onogomaoz mo ouaucoonomcc.eumuu mqm<5 63 The remaining desirable housing characteristics, with some exceptions, show similar relationships with the age of the household head. The percentage change in housing units possessing these various characteristics is less than for the two characteristics previously mentioned. Also, the percentage of units possessing the desirable housing characteristic is nearly the same for both the young and old. The age of the household head appears to be related to different measures of housing condition in a consistent pattern. The percentage of housing units possessing a desirable characteristic increases to about age 40 and then decreases. This is consistent with our previous reasoning that both old and young may experience lower levels of housing condition because they may not operate as efficiently in the housing market as household heads falling in the middle age categories. Income varies with age in this same manner and could be accounting for the variation in housing condition. Sex of Household Head This variable, female head of household, is included to test the hypothesis that the housing market as well as the credit market discriminates against women. Several causal relationships may be operative here. First, outright discrimination on the basis of sex may cause this variable to be negatively related to levels of housing condition. Or such characteristics as lower income and a higher rate of dependency may combine to cause the gross relationship between female head and measures of housing condition to be negative. A comparison between these gross relationships and the net relationships 64 should provide evidence as to the net effect of sex discrimination on the level of heusing condition. Empirical Results An estimated 17.3 percent of United States household heads were women in 1960. Summary Table 111-7 indicates that a lower percentage of households with female heads enjoyed each of the nine desirable housing characteristics than households with male heads. Each of the nine cross tabulations was tested for independence be- tween the sex of the household head and levels of housing condition. In each case this hypothesis was rejected at less than the .005 level of significance. The results presented in Table 111—7 indicate that households with female heads experience on the average lower levels of housing condition than households with male heads. However, with only one of the desirable housing characteristics, built from 1950 to 1960, does the difference exceed 15 percent. With seven of the remaining desirable housing characteristics the difference ranges between .6 and 6.6 percent. Race of Household Head A justification for the inclusion of this next characteristic, the race of the household head, could be found in the popular press. This characteristic is broken into four categories: White with a Spanish Surname, White, Negro, and Other Race which includes Indian, Japanese, Chinese, Filipino, and Other. The common presumption is that racial discrimination has existed in most markets and certainly 65 .HomH mcnmso: wcm connmnsmom mo momSmcou coon .onmemm ucoonom om .moomn mwmamm wcmmsocnum-:nuoco man an onnonnm>m meme memo; vac: -omsoc pcm mane: mcnmso: so «new mo m:0nnmnsnmn mmono Eonm oonmnsofimo one: mommncoonom omogb "condom m.5H m.wm o.mo ®.ew ©.wH m.mm H.0w m.ow m.55 0.0m 5.Nm m.wm m.m© m.ow m.cm o.ww o.mw 5.mw m.mw v.wm n5nomonmu seem an Hanon mo ncoonomv monogomsox .m.: we ommncoonom monnnnnomm cozonnx on mmooo< o>nmsnoxm oomcnsm mmonomnm no Ham: .noOHm oomcnsm nn< Enmz noumz not no Emonm onnnoonm an-nnnom ucoamnsom mcnnoom meooncnmm onoz no one coon on ommn Eona unnom noxonm no spam m on mmooo< o>nmsnoxm nonnon among a on mmooo< o>nmsnoxm menman emana none: enoo ecu no: bosom nanmnzuosnnm onoz no mEoom Xnm onEom can: monnmnnouomnmnu mcnmso: onnonnmoo one: wnonomso: mo xom an monnmnnonomnmnu manmsoz nonconom can: mung: wcnmso: .m.: powm=ooo mo omwncoonomuu.5-HHH mam<5 66 exists in the housing market. If this is true, then races other than white should exhibit a negative effect on the level of housing condition. Freeman notes that at each level of income minority races have "markedly lower quality of the housing" [22, p. 14]. This would indicate that even with the effects of income removed racial dis- crimination still has a negative effect. The net relationships estimated in Chapter IV will be examined for negative effects of races other than white after removal of the effects of other vari- ables. Empirical Results The race of the household head appeared to have strong gross relationships with levels of housing condition. In all cross tabu- lations between the race of the household head and measures of housing condition the null hypothesis of independence between the variables was rejected at less than the .005 level of significance. A summary of those estimated gross relationships is presented in Table 111-8. The primary relationships observed are the highest percentage of the housing units with the desirable housing characteristics were found among the housing units occupied by "white" household heads with "other race" next, "white with a Spanish surname” third, and "Negro” last. These relationships exist for five of the nine de- sirable housing characteristics: structurally sound, hot and cold water piped inside, exclusive access to a flush toilet, exclusive access to a bath or shower, and built from 1950 to 1960. Two of the desirable housing characteristics, structurally sound and heating 67 .noma mcnmso: use cannmnsaoa mo momomcou oomn .eHmEmm ncoonom cm .moomn onEem ocmmsocnnmncn-o:o ozn an onnmnno>m oboe meme; pnonomso: one munc: wcnmso: so «new mo moonnmnsnmu mmono Eonm woanSonmo one: mommncoonom ome£5 "oonsom m. m.w N.mw m.H m.mm m.5m m.wm o.wm 5.vm m.ov m.H5 5.5m v.m5 H.mm m.wm m.m5 m.om m.oH 5.wm m.5m H.mw v.mo w.mw N.w5 N.mm o.mo 5.om w.m5 m.ow m.mo m.Hm m.m5 o.m5 o.vm m.mw v.w© 5.vm ©.Nm m.wm H.5H fixnomonmu seem an Hench mo ncoonodv mononemso: .m.: mo ommncoonem monnnnnomu :ononnx on mmooo< e>nmsnoxm oomcnsm mmonomnm no fine: .noon oomcnsm nn< enmz none: no: no Eeonm onnnoonm on-nnn:m uceemnsom manneo: meooncnom onoz no one coon on ommn Eonm unnsm nosocm no spam m on mmooo< o>nmsnoxm nonnon amend a on mmooo< o>nmsnoxm ovnmcH pound none: onou can no: venom 5HHmn5nosnnm onoz no mEoom xnm oemcnsw oomm :mncemm m . nocno on oz onncz gun: owns: monnmnnonomnmcu manmso: eflnmnnmoo one: pnonemso: mo eomm 5n monnmnnonomnenu manmso: ponoenom nun: muncz mcnmso: .m.: penazooo mo ommucoonom--.wuHHH mqm<5 68 equipment, exhibit similar relationships to the race of the household head. However, in these cases, the households whose head is "white with a Spanish surname" have the lowest percentage of housing units with the desirable housing characteristics and the next to lowest percentage is found in households headed by "Negroes." Two other relationships are exhibited between the desirable characteristics, built from 1950 to 1960 and exclusive use of kitchen facilities and the race of the household head. Twenty-eight and seven-tenths percent of the housing units occupied by households with "white" household heads were built from 1950 to 1960, 27.9 per- cent of those with "white with Spanish surname" household heads, 26.3 percent of those with "other race" household heads, and 16.3 percent of those with "Negro” household heads. The last relationships were observed with the desirable characteristic, exclusive access to kitchen facilities. Ninety-eight and nine-tenths percent of house- holds with "white" household heads, 98.6 percent of households with "white with a Spanish surname" household heads, 97.5 percent of house- holds with "Negro" household heads, and 92.9 percent of households with "other race" household heads enjoyed exclusive access to kitchen facilities. The relationships presented in Table 111-8 indicated the households with white household heads always had the highest percent- age of households enjoying the desirable housing characteristics. With six out of the nine characteristics, households with Negro household heads had the lowest percentage of households enjoying the desirable housing characteristics. 69 Nativity and Parentage Another set of characteristics used describe the nativity and parentage of the household head: (1) native with native parents, (2) native with one foreign parent, (3) native with foreign parents, and (4) foreign with foreign parents. The hypothesis is that the closer the household head is to being foreign the less likely he will be able to operate effectively in the United States housing market. In this case, higher levels of housing condition would be associated with being native with native parents. Another hypothesis is that peeple who are foreigners or have close foreign ties may have a higher priority for housing than natives. This, in fact, would suggest the opposite relationship between nativity and parentage and levels of housing condition. The empirical results should suggest which of these forces is predominant. Empirical Results These categories describing the nativity and parentage of the household head exhibit a variety of relationships with the measures of housing condition (Table 111—9). In each of the nine cases where nativity and parentage was cross tabulated with the measures of housing condition the hypothesis of independence between the cross tabulated variables was rejected at less than the .005 level of significance. The most common pattern of relationships is exhibited with four of the desirable housing characteristics: hot and cold water piped inside, exclusive access to a flush toilet, exclusive access to a bath or shower, and one or more bathrooms. 70 .HomH menmoo: vee eOnumHemoe mo mememeeu oomn .onmemm neoonom om .momen enmEmm vemmsocnumuen-eeo ecu en oHneHnm>m eves mvme: vnocemeo: vem mane: menmson eo mnmv mo meonan:DMn mmono eonm vonmneonmo ones momeneoonom emoe5 ”oonsom 5.w e.mn w.o H.m5 mvnoeomeo: .m.: mo owmneoonom n.wm o.mm m.mm m.mm monnnnnomm coconnx on mmeoo< o>nmenoxm o.m5 o.mw o.nw o.mo oomenem mmoneene no fine; .noonm ooeensm nn< enmz none: no: no Emenm ennnoonm ennnnnsm neoeameom mennmo: m.Hm m.mm m.mm o.mw mEoonenmm enoz no oeo o.mn m.5m H.mm o.wm coon on ommn Eonm unnsm N.mm o.mm v.em 5.vw nozosm no enem m on mmooo< o>nmsnoxm m.em m.em m.em m.om eonno5 emsnm m on mmooo< e>nmenoxm m.mm m.ma n.0m e.ew menace wanna none: enoo eaa no: w.5w n.0m o.mw N.ow venom 5nnmnenoennm «.mm w.oe o.mv m.mm enoz no meoom xnm Axnomonmo comm en Hench mo neoonoev mneonme mneonea neenme mneonme emnonom canonom ewnonow oeo o>nnmz monnmnnonomnmcu menmeo: ennmnnmoo nuns emnonom can: o>nnaz can: m>nnmz can: m>nnmz emoueonee vee 5nn>nnmz 59 monnmnnonoonmeu wenmeom venoonem sen: mnne: wenmeo: .m.: venmeooo mo omeneeonoeuu.m-HHH mqm<5 71 In each of these cases, the percentage of housing units possessing the desirable housing characteristics is highest for those households whose household heads are native with one foreign parent, households whose household heads are native with foreign parents are next, house- holds whose household heads are foreign with foreign parents follow, and households whose household heads are native with native parents are last. Three other desirable housing characteristics exhibit similar relationships with nativity and parentage: six rooms or more, structurally sound, and heating equipment. In each of these three cases, the relative position of two of the nativity and parentage categories are reversed. For example, households with household heads that are native with foreign parents had the highest percentage of housing units that are structurally sound and house- holds with household heads that are native with one foreign parent are next. Two other desirable housing characteristics exhibit a different set of relationships with the nativity and parentage of the household head: built from 1950 to 1960 and exclusive access to kitchen facilities. In both cases, the highest percentage of house- holds enjoying the desirable housing characteristics were those whose household heads were native with native parents, the next native with one foreign parent, followed by native with foreign parents, and last foreign with foreign parents. The variables describing the nativity and parentage of the household head were introduced as discrete measures along a continuum from native to foreign. Two patterns of relationships with measures of housing condition were hypothesized. First, it was hypothesized 72 that the closer to being foreign a household head, the lower his level of housing condition because of a decreasing understanding of the United States housing market. The opposite relationship was also hypothesized because foreigners have different preferences for housing. Neither relationship is strongly supported by the empirical results. However, the latter hypothesized relationship may be responsible for the increased percentage of housing units with the desirable housing characteristics with categories where the household head has some foreign association. The former hypothesized relation- ship may then be resulting in a decreased percentage of housing units with the desirable housing characteristics as we move from category, native with one foreign parent, to foreign with foreign parents. At best, the relationships appear to be mixed. Metropolitan Residence in 1955 This set of characteristics is a proxy for the distance a family has moved since 1955. (The question was asked in 1960.) The set of variables represent six categories of residence in 1955: (1) same house, (2) different house same county, (3) different county same state, (4) contiguous state, (S) noncontiguous state, and (6) abroad or at sea. Certainly many instances can be found where a move from county to county was farther than a move from state to state. But in general these are assumed to represent an increasing scale of geographic mobility. It is hypothesized that geographic mobility is positively related to the household's ability to operate in the housing market and thus positively related to the level of housing condition. 73 Empirical Results The empirical results indicate that the distance moved was related to all nine measures of housing condition. With each cross tabulation between the six residence categories and measures of housing condition, the null hypothesis of independence between the variables was rejected at less than the .005 level of significance. Two distinct patterns of relationships are exhibited in summary Table 111-10. The first is the same as the hypothesized relationships where the percentage of housing units possessing the desirable housing characteristics increases with increasing geo- graphic mobility. Although the change in percentages is not always consistent between all geographic mobility categories, this general relationship exists for six of the nine desirable housing character- istics: structurally sound, hot and cold water piped inside, ex- clusive access to a flush toilet, exclusive access to a bath or shower, one or more bathrooms, and heating equipment. In only one case, hot and cold water piped inside, did the percentage of housing units with the desirable housing characteristic increase consistently over the range from same house to abroad or at sea. “However, only one inconsistency is observed for each of the other five desirable housing characteristics. The second pattern of relationships observed in Table III-10 is opposite to the hypothesized relationships previously mentioned. The percentage of housing units with the desirable housing charac- teristics decreases with increasing geographic mobility. Although some inconsistency exists, these relationships appear with two of 74 .HomH menmeo: vem eonnwneaoa mo momemeeu oomn .oHeEmm neoonom om .memmn enmemm vemmeocnumuenuoeo one en ennmnnm>m eves mvmo: vnocomoo: vem mnne: menmeo: eo mnmv mo meOnnmnznen mmono Eonm vonmneonmo one: momeneoonem omozb ”oonsom m.n m.m n.m m.n n.nm «.mm monoeomsom .m.= no omancoonoa o.na w.nm o.na m.wa e.wm N.mm mononnnoaa eoeoonn on mmooo< o>nmonoxm n.mn o.mo N.mn m.mo N.wo m.no ooacnsa moonoana no new: .noonn oomenom nn< enm: none: no: no Emonm onnnoonm enunnnem neoEmnsom wennmo: m.Hm H.Nm 5.mm 5.ww v.mw 5.mw mEooncnmm one: no eeo N.vv 0.0m 5.mv m.mv H.0m m.on coma on ommn Eon: pans: v.~m o.mm v.00 0.0m N.5w o.©w no:o:m no seem a on mmeoo< e>nmsfiuxm H.mm N.mm m.nm 0.0m 5.mw v.5w nonnok Amen: m on mmeoo< e>nwsHuxm c.5m m.om H.mm o.Nm m.mw m.5m ovnmeH vomnm none: vHou vem no: m.mw o.ww o.mw N.mm N.Hw m.mm venom xfiamnzuosnnm w.wH v.vm v.wm N.Hm N.wm H.vv one: no mEoom xnm mxnowonwu comm en Hmn05 mo neooneev mom onmnm onmnm neeeou no no msoemnneoo enmnm osmm oEmm omoo: monnmnnonoonmcu menmso: onnmnnmoo vmonn< -emz msoemnneou .5nesou .omzo: oeom . . . . econooona neononena mmmn en eoeovnmom emunnomonnez 5n monnmnnonomnmcu wenmeo: venoonom :nn: mnne: menmeo: .m.: veneeooo mo emmneeunemua.oH-HHH mqm<5 75 the desirable housing characteristics: six rooms or more and ex- clusive access to kitchen facilities. A third set of relationships appears with the desirable hous- ing characteristic, built from 1950 to 1960. The changes in percent- ages are too varied, however, to represent a distinct pattern. The existence of two opposite patterns of relationships is demonstrated in Table III-10. This means that depending upon the measure of housing condition chosen, geographic mobility can be shown to have a negative or positive relationship to housing condition. Occupational Classifications Next, a set of eleven variables are used to denote different occupational classifications. A list of the specific occupations which constitute these aggregate classifications appear in several publications [24, 35, 36]. The aggregate classifications used in this study are defined here. White collar workers encompass: (1) professional, technical, and kindred workers; (2) managers, officials, and preprietors, except farm; (3) clerical and kindred workers; and (4) sales workers. Blue collar workers encompass: (1) craftsmen, foremen, and kindred workers; and (2) operatives and kindred workers [36, pp. 41-47]. Other categories used are: farmer, farm manager, farm foreman, farm laborer, farm service worker, service worker, laborer, occupation not reported, and no occupation. These variables which also appear in Spurlock's work [17, pp. 21, 33] are used as a crude proxy for the taste of the household regarding housing. Some of the occupational groupings may be too broad to approximate tastes. However, due to the costs in computer time of a large number of II 76 observations and variables, we chose to use aggregate classifications except for occupations in which we are most interested. As well as the effects of preferences of households, the gross relationships of occupational classification variables will include some effects of income and educational differences on levels of housing condition. Empirical Results The estimated gross relationships between occupational classifications and measures of housing condition are presented in Table 111-11. In each of the nine cross tabulations between occu- pational classifications and measures of housing condition, the null hypothesis of independence between the variables was rejected at less than the .005 level of significance. The various occupational groups show considerable variation in the percentage of housing units possessing the desired housing characteristics. For example, the largest range of percentages was found with the housing characteristic, heating equipment. Eighty- three and nine-tenths percent of the households in the white collar group enjoyed the desirable types of heating equipment and only 20.0 percent of the households in the farm labor group had this heating equipment, a 63.9 percent range from minimum to maximum. Five other desirable housing characteristics have a range which varied from 51.2 to 56.1 percent in magnitude. The range for six rooms or more is 34.7 percent, For built from 1950 to 1960, the range is 27.0 Percent, and for exclusive access to kitchen facilities the range is ‘the smallest, 3.5 percent. ......a— no}. 12‘ on... a; o. 5...... ~....\ 5: m 7....J. —..,- hyepu ~ any. n— a .-.S .r.. a — n-~ aJ~5 ~ ua.-flv- .7..N~ ween ~5-~IJIJNV IL nu ova-..b5~nW:Je~nULl I o N N] NNN MN.NN~F\N 77 .H0mH menmao: vem eonnmnsmoe mo momemeou 000: .enmEmm neoonom on .mommn onEmm vemmeocnumuenueeo oen en enpmnnm>m eves mvmo: vnoeemeo: vem mnne: menmeo: eo onev mo meonnmnenmu mmono Eonm vonmnoonoo one: mowmueoonoe omogn “eonsom v.H 0.0 m.m 0.0 5.5 H. 0.0 H.mm n. v.mm mvnogomso: .m.: mo ommneeonoa 0.00 0.00n H.00 0.00 0.00 0.00: 0.50 «.00 0.00: 0.00 monnnnnomm eogonnx on mmeoo< e>nmenoxm 0.0m 5.0m 0.00 0.00 0.00 0.00 5.00 0.50 H.mm 0.mw eomenam mmonomnm no nae: .noon oomen30 nn< Enos none: no: no eoenm onnnoonm enunnnsm neoemnscm mennmo: 0.00 0.00 0.50 m.m5 0.05 0.05 m.~w 0.5m m.n0 0.00 msooncnmm enoz no eeo w.~n H.NH 5.0m 0.0H N.0H 0.5m 5.0m n.0m n.5n 0.00 000a on 000H Eonm ennem v.0e 0.00 v.n5 0.55 0.nw 0.N0 H.00 0.00 5.00 H.50 ne:o:m no seem a on mmooo< o>nmenoxm 0.0a 0.v0 0.05 0.05 5.00 m.0m 5.mm 0.00 m.n0 0.50 nonnon :mznm a on mmooo< o>nmnnexm 5.vv 0.50 0.05 N.m5 v.00 0.Nm ~.0w 0.H0 m.n0 v.00 ovnmeH voene none: vnou veo no: 0.Nv H.N5 5.00 0.N5 n.05 0.00 0.05 w.mm 0.50 0.00 venom 5nnmnenoennm m.m~ «.00 5.0m 0.3. wém 0.2“ 0.0m 0.3 0.50 5.: one: no 280: xnm 55nomoumu :omm en Hmuon mo neooneav no a eOnn eon ems vonnom no o new no 0 men mane omno emu“ noenmm mnonopoq -mm3ooo u>n0 nonom uom own 0 -memz o Mn 0 wenwwo. on mnnmw oz 0 Snow noz H0 and: n.:2 . : an . 0 vme: vnonemeo: one mo moono Hmeonnem unooo 59 monnmnnonomnmau menmeo: venoenem nun: mane: menmeo: .m.0 venmooeo mo emeueoonom...g HHH a mqm<5 78 The summary relationships reveal some consistency in the way occupational groups relate to measures of housing condition. For six of the desirable housing characteristics, households whose heads are white collar workers have the highest percentage of housing units with those housing characteristics. In other words, the occupational group, white collar, ranked first for six of the desirable housing characteristics. Farm managers ranked second for five characteristics. Blue collar workers ranked third for five characteristics. Not re- ported ranked fourth for five characteristics. Farm foreman exhibited mixed results. Service workers ranked sixth for four characteristics. No occupation ranked seventh for six characteristics. Laborers ranked eighth for seven characteristics. Farmers ranked ninth for six characteristics. Farm laborers ranked last for seven characteristics. Only two of the desirable housing characteristics, exclusive access to a bath or shower and one or more bathrooms consistently exhibit this pattern or relationships with the occupational groups. Thus the relationships between these occupational groups and housing condition will vary depending upon the measure of housing condition used. Type of Tenure Variables are included to describe the type of tenure the household enjoys: (1) owned, (2) rented, or (3) no cash rent. We hypothesize that the person who rents is less likely to place as much importance on the housing unit as the owner because of the difference in preperty rights. Thus owners on the average should enjoy a higher level of housing condition than renters. The household 79 that does not own and pays no cash rent for their housing unit is expected to place the least importance on their housing units of the three groups of households and thus experience the lowest level of housing condition. Empirical Results The estimated gross relationships between types of tenure and measures of housing condition are presented in Table III-12. For each of the nine cross tabulations summarized in this table, the null hypothesis of independence between the variables was rejected at less than the .005 level of significance. For seven of the nine desirable housing characteristics pre- sented, the highest percentage of housing units with those charac- teristics was among the owned units, the next highest percentage was among rented units, and the lowest percentage was in the category no cash rent. This result agrees with our hypothesized relationships between type of tenure and housing condition. Owned housing units tend to have a higher condition level than rented units which have a higher condition level than rented units for which no cash rent is paid. For two of the desirable housing characteristics, six rooms or more and exclusive access to kitchen facilities, a higher percent- age of housing units possessing the desirable housing characteristics are found in the category, no cash rent, than in the category, rented. These are the only two exceptions to the pattern of relation- ships discussed earlier. .H0mu menmno: vee eOnnmnseom mo mememeou 000a .enmemm neoonem 0N .momen emmamm vemmooenumuenuoeo o:n en onnmnnm>m evme mvmo: vnocemeo: veo mnnee wenmeo: eo mnmv mo meonnensnmn mmono Eonm vonmnsonmu one: mewmneoonom omoen ”oonsom 80 0.0 H.0m 0.n0 mvnoeomeo: .m.: mo ommneoonem 0.50 0.00 0.00 monnnnnomm eononnx on mmeoo< o>nmenoxm N.0m 0.00 5.N5 oomenam mmonoena no Ham: .noon oomensw nn< enm: none: no: no Emenm onnnoenm ennnnnem neeEmneom mennme: 0.00 5.nm v.00 msoonnnmm one: no oeo 5.0a m.0n 0.00 000a on 000a Eon: nnnem 0.00 n.00 . 0.H0 no:o:m no :nom m on mmooo< o>nmnnoxm «.50 5.00 0.H0 nonno5 seen: 0 on mmeoo< o>nmenoxm N.0m 5.00 m.H0 evnmeH vomne none: vnou vem no: 0.00 m.m5 0.00 venom nanmnenoennm N.Hm H.0H 0.00 enoz no mEoom Xnm finnomenmu oneeon Loom en Hmno5 mo neeonemv nee: :mmu oz voneom vee:0 monnmnnonomnmeu menmeo: enamnnmea oneeoh mo e055 52 monnmnnenomnmnu menmoo: venoonem :nn: mnne: menmeo: .m.: venmeooo mo emeneeonemu-.NH-HHH mqm<5 81 Education of Household Head The educational level of the household head is thought to be another characteristic which would likely be related to housing con- dition. We hypothesize that as education increases, both the desire for housing and ability to function in the housing market increase. Also, we believe that low levels of education are positively associ- ated with a number of other socio-economic and locational character- istics almost all of which have a negative effect on the level of housing condition. For example, education is thought to be positively related to income, old age, rural location, etc. This would result in a positive gross relationship between education and levels of housing condition which is larger than the positive net relationship where the effects of other variables are removed. Empirical Results The estimated gross relationships between the education of the household head and measures of housing condition are presented in Table III-l3. For each of the nine cross tabulations summarized in that table, the hypothesis of independence between the variables was tested. For eight of the nine cross tabulations, this null hypothesis was rejected at less than the .005 level of significance. For the last cross tabulation, exclusive access to kitchen facilities and education of the household head, the null hypothesis was rejected at less than the .05 level of significance. The basic pattern of relationships revealed in the summary table is as hypothesized. Education of the household head and the 82 .HonH «enmeo: vee eonneneeom mo monomeou 000a .eneaem neoonee 0N .momen unmade vemmeonn-euenueeo een en ennonne>u eves nveo: vnoseneo: vea anneo menmso: eo onev mo meonnanonen «mono Bonn venensonoo one: momaneoonee omonh “oonaom mvnoeemso: .m.: mo eueneoonom n.0 0.0 n.m o.n~ 0.0 0.0n 0.5: 0.0 0.n 0.0 n.~ ~.ma n.00 n.0o 0.00 0.00 n.0m 0.00 0.00 n.0m 0.00 0.00 nononnnooa eenunnx on mmouo< o>nm=noxm n.00 0.00 0.00 n.0n n.~n m.n0 0.00 0.00 n.n0 n.00 n.00 ooocnoa oneneenm no anus .noonu oooenem nn< anus none: no: no aaonm onnnoonm an-nnnom neoamnacm mennee: 0.0a 0.na n.00 n.00 0.:0 0.00 0.~0 0.0n 0.00 5.00 0.50 maoongoom ono: no 0:0 n.n0 n.00 n.nn n.00 0.0N 0.n~ 0.0: n.0n n.0n 0.0n 0.0 o00n on omen aona unnam n.0o n.no 0.n0 0.00 5.~0 ~.oo 0.00 0.05 n.~n n.00 5.~0 nosogm no soon a on mmeou< e>nmenoxm 0.00 n.na 0.nm 0.00 0.00 0.nm 0.00 0.00 0.0n n.00 0.n0 nonnon gonna a on mmeoe< e>nmsnoxm 0.00 0.00 0.00 0.00 n.0a 0.no 0.n0 0.00 0.0n 0._0 «.00 oonmen ooana none: vnoo vee no: N.na «.00 0.00 0.00 n.n0 0.00 5.00 0.0n 0.00 0.00 0.00 venom nnnonsooonom 0.00 0.00 0.00 n.00 0.00 0.00 n.00 0.n0 n.0n o.0~ n.- ono: no oaooa xnm annouoneu neeOnnaoevm seam en Hench mo neoonoev onoz no 0 0-n 0 0 N.n 0 n 0.0 0-n 00oz 0 .noo .noo .noo .0 sun: .0 sun: .0 sun: .aonm .aonm .aonm .aonm vac: vnoeomoo: mo eonneoevm mennmnnonuenoeu uenmso: enpunnmoo “OwuuwHDHUQHdg "pawn—do: VOHUOflOm £HH3 bOflOHGU GOMHQUq—fim AUG” Gm ”303 VHOAOWSO: NO ONCHGOONOAII.M~IHHH 043%. 83 level of housing condition are positively related. Only the desirable housing characteristic, exclusive access to kitchen facilities exhibits a different relationship which does not have any consistent pattern. A substantial exception to the basic pattern appears between the edu- cation categories of none and elementary 1-4. In seven cases, out of nine, the category, elementary 1-4, has a lower percentage of housing units with the desirable housing characteristics than the education category, none. Other than the two exceptions mentioned, there are several inconsistencies where the percentage of housing units with the desirable housing characteristics is higher for an educational category which represents less education than another. However, these inconsistencies are few. The basic pattern of a positive relationship between housing condition and educational level of the household head seems to prevail. Household Income Income is a variable which economic theory would tell us is positively related to levels of housing condition. In this case, we have considered income of two distinctly different types of house- holds: those consisting of unrelated individuals and those con- sisting of families. Gross relationships between income and levels of housing condition are examined separately because it was felt that the relationships between household income and housing condition would be different for these two groups. Also, notice that the income figure used was household income or the sum of all income for the household. This was felt to be the relevant figure because the household head may not be the major income recipient and thus 84 household expenditures may depend on another member's income. Having no means of determining which member's income is the primary source for the household budget, we chose to sum total income for all house— hold members feeling this would constitute a more relevant measure than income of the household head. Schaeffer and Edwards noted that this total income figure seemed to be a better explanatory measure: "When the effect of all income sources are added, the correlation is increased to .85 from .82 for heads of family income only" [15, p. 12]. Empirical Results The estimated gross relationships between household income and measures of housing condition for two types of households, families and unrelated individuals, are presented in Table III-l4. For each of the cross tabulations summarized here except one, the null hypothesis of independence between the variables was rejected at less than the .005 level of significance. For the cross tabu- lation between the housing condition measure, exclusive access to kitchen facilities and household income for unrelated individuals, the null hypothesis could be rejected at less than the .1 level of significance. The results presented in Table III-14 support the hypothesis that household income and housing condition are positively related. For all of the desirable housing characteristics except exclusive access to kitchen facilities, the percentage of housing units possessing the desirable housing characteristics increases as income 85 0.00 0.00 0.50 «.00 5.50 0.50 0.50 «.00 0.00 0.00 0.H0 n.05 0.00 5.00 nmennneem 0.00 0.00 5.00 0.00 5.00 0.00 0H00 0.00 «.00 0.00 0.00 5.05 0.05 0.n0 omnoavn>nven venenonen maoonenom onoz no eeo 0.«0 5.00 v.Hv m.nv 0.00 H.0n 0.50 0.00 0.00 0.0« n.0« 0.0« «.5H 0.0H nmennnsam 0.00 n.vv 0.0« 5.5a 0.5m 0.n« 0.0« «.«« 5.0m n.0H 0.«H 0.0n 0.0“ 0.0n . emneevn>nven venononea 000a on 000a aonm nnnsm 0.00 «.00 «.00 v.00 n.00 0.50 5.50 0.00 0.00 0.00 0.00 0.55 0.50 ”.00 pmonnnsam 0.00 «.00 5.00 0.00 0.00 0.00 0.00 H.00 5.00 «.«0 0.00 0.55 0.05 «.00 omnaevn>nven venmnonez nozoem no enmm m on mmeou< e>nmenoxm 0.00 0.00 0.00 0.00 0.00 0.00 ”.00 5.00 «.00 m.H0 0.00 0.05 0.05 0.n0 pmennnamu 0.00 «.00 5.00 0.00 0.00 0.00 0.00 «.n0 «.50 0.00 «.00 0.05 0.05 0.00 unneevn>nven venenonea nonnon amen: e on mmooo< o>nmenoxm 0.00 «.00 «.00 0.00 0.00 «.00 n.00 0.00 0.00 0.«0 0.00 0.05 0.00 0.00 amonnnaum «.00 n.50 0.00n 0.00“ 0.00 «.00 0.50 0.50 5.00 «.«0 0.00 0.50 0.00 0.«5 omneevn>nven venenonea 000000 oonna none: onoo 0:0 000 0.50 ”.00 H.00 0.00 0.00 0.00 «.«0 0.00 0.00 0.00 0.05 H.«5 «.00 «.00 nmennneem 0.00 «.00 0.00H H.00 0.00 0.00 5.00 n.00 0.00 0.05 «.00 0.05 v.V5 v.00 emnuavn>nven vonmnonea vesom nnnenenesnnm 0.05 «.00 0.nm 5.0m n.0v 0.00 v.nv «.00 0.00 «.n» 5.0« 0.5« 0.0« 0.5« nmonnnaum 5.n0 0.«0 0.0« 0.Hn 0.0a 0.0« n.0n 5.0“ 0.0" 0.00 0.0” 0.0n 5.0a «.0n omneevn>nven venonenez one: no maoom xnm annoueneu oaoeen zoom en nonon mo neeoneav man man «H0 Hag on; 00 0a 50 00 00 «A 00 «a n4 «nu n00 000 00 00 50 00 00 v0 00 «0 no 00 mennmnnonuuneeu unannoo mo nveemaonb en eaooen vnozemeo: menmao: ennennaeo munnmnnonuanueo uenmeo: venoenem sun: monnoueneu oaooeH enunoomm vee vnoeemeo: mo e005 An mvnonemeo: mo oneneeunoeus.vn-nHH mam<5 nail-Lu unseenvll .h.~INNN m~<-flAh 86 .H000 menmeo: veo connonemom mo momnmeou 000a .onmsom neeenoa 0« .monon enmeom veomaogn-o-en-oeo enn en onnmnno>o eves mvoo: vnonomeoe veo annee menmso: eo onov mo meonnonsnon moonu Eonm venoneonoo one: mouoneeonon omozn ”eonsom .monnneom mo vomoesoo mvnonomeo: 0 .mnoevn>nven venonone: mo vomomsoo mvnozomaozo .on Hence no eoen moon meooe 0 .eoen nonoonu memos 0 "onoz 0.00H 0.00 0.00H 0.00 0.00 0.00H 0.00H 0.00 0.00 5.00 5.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.«0 0.00 0.«0 «.00 0.~0 0.«0 0.00 0.«0 5.00 0.00 0.00 5.00 N.00 0.00 0.05 5.H5 0.v0 0.00 0.00 0.Hv 0.00 0.«0 v.00 5.00 0.05 «.«0 0.00 0.00 0.55 0.55 5.95 0.00 9.00 0.00 0.00 Annouonou oaooeH zoom en Hanan mo neounomv amonnnaan annooon>n0=n 00000000: mvnogomso: .0.= mo emoneoonom 0monnnaoa amnoaon>nocn 00000000: monnnnnoon eoeonnx on mmooe< o>nmenoxm pmonnnaom omnoevn>nven venononea ouoensm mmonomnm no nae: .noonm oooenem nn< anus none: no: no aoonw unnnuonm en-nnn=0 neoannscm wennoo: nu: «a: mag On: 04 04 54 04 mg v; we N4 H4 000 ~00 nnu one 00 00 no 00 00 00 00 no no 00 unannoa mo mveomeonn en oeooeH vnonomso: monnmnnonoonoeu menmeo: onnonnmoo ooocnocoo--.en-nnn 000o vnonomso: veo unne: uenmso: eo onov mo meonnonsnon «mono sonm venonaenoo one: memoneoonom omoen "oonsom eves mvoo: .on Hence no: memos .o.e .on Hence no eoen omen meooe q .eoen nonoonu meooa 0 "onoz «.0n 0. 0.« 0.0 v. 0.5 «. n.0n «.0 0.0n n.« 0.00 mvnoeemoo: .0.0 no ouoneeonoe n.00 0.00 0.00 0.00 5.00 0.00 0.00: 0.00 0.00 0.00 0.00 0.50 mennnnneom eoeonnx on mmooo< o>nmanexm 0.00 0.«0 5.00 H.00 5.00 5.n5 0.00 n.00 0.00 0.05 «.00 0.05 oeoenam mmenomnm no gnu: .noonm oooeneu nn< ano: none: no: no soonm onnnuenm en-nnn:0 neoamnoum mennoe: 0.05 n.00 0.05 0.00 n.05 5.00 «.00 0.50 «.00 5.00 0.00 0.00 maoonenom enoz no eeo 5.0M 5.0« 0.0« 0.00 5.0n «.00 v.0n 0.00 0.0« 0.n0 «.0H 5.0« 000: on 000n aonm nnnem 0.00 0.00 0.05 5.00 5.«5 0.00 «.50 0.00 0.00 0.00 5.50 0.50 no:o:m no anon a on mmooe< o>nm=nexm v.00 0.n5 0.05 0.50 0.05 n.n0 «.50 5.00 0.00 «.H0 «.00 0.00 nonnon amen: a on mmoee< e>nmenuxm 0.00 5.00 «.05 0.00 v.n5 0.00 5.00 0.00 0.00 n.n0 0.00 v.n0 ovnmen 0002 none. Boo 05 no: 0.55 0.00 0.00 0.55 n.00 0.«0 «.00 ”.00 0.05 0.00 0.00 0.00 venom Annonenuannm 0.00 «.50 0.50 0.50 0.00 0.50 5.00 0.00 0.50 0.00 0.50 «.0« one: no 03000 000 Annomenou neeoveemeo zoom :0 Hanan mo neounemv neeveemon ane.e 00 «a 05.n0 0.00 0«.n0 0.00 05.0 0.0 0«.0 0 nn< 00 «0 05.00 0.no 0«.n0 0.n0 05.0 0.0 0«.0 00 annex neeoveoqeo monnmnnoneonoeu menmeo: enponnmeo mennmnnenuenecu uennao: veneonom Ann: monnouonou neeeveomon enunoegm en mvnoeeneo: mo emeneoonemua.mH-HHH mqm<5 90 Summary and Conclusions In this chapter the estimated gross relationships between socio-economic and locational characteristics of the occupants and various measures of housing condition have been presented. Some sets of these characteristics appeared to explain more variation in hous- ing condition than others. That is, some sets of characteristics exhibit a larger range of percentages of housing units with selected housing conditions than other sets. The occupational classifications contained the largest range for six of the selected housing charac— teristics. Household income, educational level of the household head, size of place, and tenure generally exhibit a slightly smaller range of percentages than the occupational classifications. Location within an urbanized area, sex of the household head, and the metro- politan residence in 1955 appear to explain the least amount of vari- ation in housing condition. The four other sets of socio-economic characteristics are between these two extremes. They are listed here from the set with the strongest estimated relationships with housing condition to the weakest: dependency ratio, race of house- hold head, region of the United States, and the nativity and parentage of the household head. This ranking of socio-economic and locational characteristics as to the strength of their estimated gross relation- ships with housing condition was done through comparing the range of percentages across all of the selected housing characteristics. The ranking may not fit any particular selected housing character- istic. However, the generalization does present some information on the characteristics which seem most highly related to housing condition. 91 The gross relationships estimated and presented in this chapter will be used again in Chapter V when they are compared to the net relationships. The estimated net relationships between the socio-economic and locational characteristics of the occupants and various measures of housing condition will be presented in the next chapter. CHAPTER IV NET RELATIONSHIPS BETWEEN SOCIO-ECONOMIC AND LOCATIONAL CHARACTERISTICS AND HOUSING CONDITION: PREDOMINANT INFLUENCES Introduction How do income, sex, race, age, and education relate to levels of housing condition? This is the type of question that will be asked and answers suggested in this chapter and the next. The previous chapter dealt with the gross relationships between thirteen sets of socio-economic and locational characteristics and housing condition. In this chapter and Chapter V we present the estimated net relation- ships between these same characteristics and housing condition. Several procedures are employed in order to present these estimated net relationships. First, the set of socio-economic and locational characteristics used in Chapter III are included as re- gressors in several multiple regression models. Each of these re- gressions with a common set of regressors, has a different measure of housing condition as the regressand. This procedure is used to estimate the net relationships with each of the measures of housing condition just as cross tabulations or contingency tables were used to estimate the gross relationships in Chapter III. Second, this same set of regressors are used in a regression model with the INDEX 92 r0 r1 93 as regressand. This procedure is used to estimate the net relation- ships between the socio-economic and locational characteristics of the occupants and the measure of housing condition, INDEX, explained in Chapter 11. Each of these models is presented separately. The third procedure involves examining the INDEX for weight sensitivity. Twenty different sets of weights are used on the components of the INDEX, while INDEX 1 through INDEX 20 are used as regressands in twenty regressions. The same set of socio-economic and locational characteristics as are used in the other models in Chapter IV are used as regressors here. The resulting parameter estimates are examined to determine if the INDEX is weight sensitive. The pre- sumption is that the estimates should remain relatively constant if the INDEX is not to be judged weight sensitive. The specifics of these three procedures will be developed as the chapter proceeds. This chapter is organized around the three research pro- cedures just discussed. The first section includes specification of the functional form of the socio-economie and locational character- istics used in the regression models. The second section includes a presentation of the net relationships estimated in the first pro- cedure. Each model is examined in total for the relative importance and direction of relationships between the thirteen sets of socio- economic and locational characteristics and each measure of housing condition. Section three includes the estimated net relationships between the regressors previously used and the INDEX. The fourth section includes an examination of the INDEX for weight sensitivity. This is followed by the summary and conclusions regarding the net 94 relationships between selected socio-economic and locational charac- teristics and housing condition. This chapter includes specification, presentation, and dis- cussion of each model indicated above. If the reader is not inter- ested in each individual model, the estimated net relationships are presented in Chapter V in a format similar to that used for the esti- mated gross relationships in Chapter III. That is each of the models is divided to present the estimated net relationships between each set of socio-economic and locational characteristics and the various measures of housing condition. Model Specification This section includes specification of the functional form of the socio-economic and locational characteristics to be used as regressors in the regression models of this chapter. The information used in specifying these independent variables comes from several sources: (1) the estimated gross relationships presented in Chapter III, (2) net relationships estimated using ”abbreviated regression models," and (3) previous studies. Abbreviated Models The abbreviated regression models were used only to obtain information on the functional form of predetermined variables. Thus they are not presented in detail. They differ from the models used in this chapter in several ways. First, fewer socio-economic and locational characteristics are included. Second, the characteristics included are described by a set of binary variables. Third, the 95 models are estimated with only a portion of the sample ultimately used. The binary regressands used in these eight abbreviated re- gression models are presented below. Y = 1 if a telephone is available 0 otherwise Y = 1 if household has exclusive access to kitchen facilities 0 otherwise Y = 1 if hot and cold water are piped into the housing unit 0 otherwise Y4 = 1 if there is exclusive access to a flush toilet 0 otherwise Y5 = 1 if there is exclusive access to a bath or shower 0 otherwise Y = 1 if the unit was built from 1950 to 1960 6 0 otherwise Y7 = 1 if the unit contained these four better types of heating equipment: (1) built-in electric, (2) steam or hot water, (3) warm air furnace, and (4) floor, wall, or pipeless furnace. 0 otherwise Y8 = 1 if the unit has eight rooms or more 0 otherwise The sole purpose of these models was to provide information about the functional form of variables which would be used in the final models. Variables which would be continuous in the final models were broken into intervals and described with binary variables. The estimated regression coefficients were then plotted to determine the functional form to be used in the final models. Information from this source was used to specify three sets of socio-economic and locational characteristics: size of place, educational level of household head, and household income. 96 Predetermined Variables The functional forms of the sets of socio—economic and locational characteristics will be presented in the same order as these characteristics were introduced in Chapter III. The rationale for their inclusion and hypothesized relationships which was included in Chapter III will not be repeated here. Regions of the United States The regions of the United States are included as binary variables. X1 = 1 if the household resides in the Northeast 0 otherwise X = 1 if the household resides in the North Central region 0 otherwise X = 1 if the household resides in the South 0 otherwise X = 1 if the household resides in the West 4 0 otherwise X1, X2, X3, and X4 represent an all inclusive set and the regression model has a constant term. Therefore X was dropped so 2 the model could be estimated [36, p. 19]. Size of Place The next set of variables, size of place, contain a mixture of discrete and continuous variables. X5 = 1 if the household is rural farm 0 otherwise X6 = 1 if the household is rural nonfarm 0 otherwise 97 X7 = 1 if the household is in an urban territory outside of places 0 otherwise X = the logarithm to the base 10 of the population of the household's place of residence The population of the household's place of residence was included as a logarithm to the base 10 after examining the parameter estimates from the abbreviated models. Binary variables were used to describe various population intervals. When the parameter esti- mates were plotted on log paper, the size of place variables appeared to have a log linear relationship with each of the binary regressands. Intuitively, these relationships appear plausible. In fact, one would expect an addition of 5,000 population to a place of 10,000 population to have a greater affect on the functioning of the housing market than the same addition to a place of 50,000 population. The log linear specification will allow for this type of relationship [36, pp. 19, 20]. Location Within Urbanized Area The two residence categories distinguished within an urbanized area are, in a central city and in the remainder of an urbanized area. They are represented by binary variables. X9 = 1 if the household resides in a central city 0 otherwise X10 = 1 if the household resides in the remainder of an urbanized area 0 otherwise These residence categories are determined only for residents of urbanized areas. Only X9 was included in the models to examine the effects of being in the central city on housing condition. X10 98 was omitted because of the belief that the additional information it could provide was more completely provided by the size of place variables [36, p. 21]. Age of Household Head The age of the household head is described by three continuous variables. x11 X12 - the age of the household head squared the age of the household head X13 - the age of the household head cubed The age of the household head was included as a cubic function after plotting the percentages estimated in the cross tabulations of Chapter III. An examination of Table III-6 reveals a pattern of relationships between age and housing condition which, it was be- lieved, could be well represented by the cubic form [36, p. 6]. Sex of the Household Head The sex of the household head is described by two binary variables. X = 1 if household head is male 14 . 0 otherwise Xls = 1 if household head is female 0 otherwise Because a constant term is included and X14 and X15 form an all inclusive set, X14 is dropped to provide for estimation of the parameters [36, p. 22]. () L.,c 99 Race of Household Head The race of the household head is represented by four binary variables. X16 = 1 if the household head is white with a Spanish surname 0 otherwise X = 1 if the household head is white and X = 0 l7 . l6 0 otherwise X18 = 1 if the household head is Negro 0 otherwise X19 = 1 if the household head is Indian, Japanese, Chinese, Filipino, or other or X = X = X = O 16 17 18 These binary variables form an all inclusive set. Also, a constant term is included in the models. Thus X was dropped to 17 allow for estimation of the models [36, p. 14]. Nativity and Parentage The nativity and parentage of the household head are described by four binary variables. X20 = 1 if the household head is native with native parents 0 otherwise X21 = 1 if the household head is native with one foreign parent 0 otherwise X22 = 1 if the household head is native with foreign parents 0 otherwise X = 1 if the household head is foreign 23 . 0 otherwise Because a constant term is included and X20, X21, X22, and X23 form an all inclusive set, X was dropped to provide for esti- 20 mation of the parameters [36, p. 25]. 100 Metropolitan Residence in 1955_ The metropolitan residence of household head in 1955 is described with binary variables. x24 x25 26 27 28 X29 x24 1 0 1 l 0 if the household head occupied the same house otherwise if the household head resided in the same county but a different house otherwise if the household head resided in the same state but a different county otherwise if the household head resided in a contiguous state otherwise if the household head resided in a noncontiguous state otherwise if the household head was abroad or at sea otherwise is dropped from this set of variables to provide for estimation of the parameters [36, p. 36]. Occupational Classification Eleven occupational categories of the household head are described with binary variables. x30 x31 x32 x33 34 — 1 0 if the household head is a farmer otherwise if the household head is a farm manager otherwise if the household head is a farm foreman otherwise if the household head is a farm laborer otherwise if the household head is a farm service worker otherwise 101 x35 = 1 if the household head is a white collar worker. (This category includes: (1) professional, technical, and kindred workers; (2) managers, officials, and proprietors, except farm, (3) clerical and kindred workers; and (4) sales workers.) 0 otherwise X36 = 1 if the household head is a blue collar worker. (This category includes: (1) craftsmen, foremen, and kindred workers; and (2) operatives and kindred workers.) 0 otherwise X37 = 1 if the household head is a service worker. (This category includes: (1) private household workers, and (2) service workers, except private household.) 0 otherwise X = 1 if the household head is a laborer 38 . 0 otherwise X39 = 1 if the occupation of the household head is not reported 0 otherwise X = 1 if the household head has no occupation 40 . 0 otherwise This set of variables is all inclusive so X is dropped to 39 provide for estimation of the parameters [36, pp. 40-47]. Type of Tenure Three types of tenure are described by three binary variables. X = 1 if the housing unit is owner occupied 41 . 0 otherwise X42 = 1 if the housing unit is renter occupied and the renter pays cash rent 0 otherwise X43 = 1 if the housing unit is renter occupied and the renter pays no cash rent 0 otherwise For most of the models presented, X42 has been dropped to provide for estimation of the parameters. However, in some models, 102 both X42 and X43 have been dropped due to an oversight. Caution must be exercised in comparing the parameter estimates for X4 between 1 models where X42 has been dropped and those where X and X43 have 42 been dropped. Where X42 has been dropped, the parameter estimate for X41 describes the difference between the effects of X41 and X42 on the regressand and the parameter estimate for X43 describes the difference between the effects of X and X4 on the regressand. 43 2 In models where both X4 and X43 have been dropped, the parameter 2 estimate for X41 describes the difference between the effects of X41 and the combined effects of X42 and X43 on the regressand [36, p. 69]. Education of Household Head The educational level of the household head is described by two variables. X44 = the number of years of formal education if less than or equal to 10.5 years 10.5 otherwise X4S = the number of years of formal education if greater than 10.5 years 0 otherwise This functional form was chosen after examining parameter estimates from the abbreviated regression models discussed earlier. In these models, binary variables were used to describe the house- hold head's years of formal education. The parameter estimates were then plotted to obtain information on the functional form of the continuous relationships between years of education and the desirable housing characteristics used as regressands in the abbreviated models. Two distinct patterns emerge. The first is approximately linear. 103 As the educational level of the household head increased, the proba- bility of the housing unit containing the desirable housing character- istic increased linearly. The second pattern included two linear portions. Up to 10.5 years of formal education the probability, that the desirable housing characteristic was present, increased linearly. After 10.5 years of formal education, the probability increased linearly but at a smaller rate. The specification used here allows for this kinked relationship and for the one without the kink. The data on education are included in the Census as discrete categories, some covering more than one additional year of formal education. The approximate midpoint of these categories was chosen as the value of the continuous variables used here, X44 and X45. These are the values used for the various categories: Category Value None 0 Elementary 1-4 2 Elementary 5 or 6 5 Elementary 7 6. 7 9 U'lUl Elementary 8 High School 1 or 2 High School 3 10.5 High School 4 11.5 College 1—3 13.5 College 4 15.5 College 5 or More 16.5 Specific definitions of the census categories may be found in the technical documentation of the sample [36, pp, 37, 38]. Household Income Household income is described by five variables, three continuous and two binary. 104 X I 46 - the logarithm to the base 10 of household income if the household is composed of unrelated individuals 0 otherwise X47 = 1 if the household is composed of a family 0 otherwise X48 = the logarithm to the base 10 of household income if the household is composed of a family 0 otherwise X49 = 1 if the household is composed of a family or families and unrelated individuals, i.e., if the household is "mixed" 0 otherwise xso the log of household income if the household is ”mixed” 0 otherwise All negative income is given the value of $4.50. Also, income that is greater than $7,000 is given the value of $7,000. The evidence for this specification comes from both the contingency tables of Chapter III and the abbreviated regression models. In the abbreviated regression models, household income was entered as a series of binary variables with all types of households lumped together. The plotted parameter estimates revealed relation- ships between household income and desirable housing characteristics which could be approximated by the log-linear functional form. An examination of the contingency tables of Chapter III also revealed relationship that could be approximated by the leg-linear functional form. All types of households were not lumped together in the cross tabulations. Separate contingency tables were constructed for households composed of unrelated individuals and for households com- posed of families. No cross tabulations were constructed for house- holds that are a mixture of these first two types. An examination of summary Table III—14 reveals that the two types of households 105 have different levels of housing condition at any one income level. A plotting of these relationships also suggests that the relationships between household income and levels of housing condition have differ- ent slopes for the two types of households. Representing household income by a mixture of binary and continuous variables allows for the suspected differences in relation- ships. Households are divided into three types, households composed of unrelated individuals, those composed of families, and a mixture of the first two. It is assumed that the relationship between house- hold income and housing condition is log-linear, but different for each type of household. The two binary variables, X47 and X49 allow for differences in the intercepts of the three relationships, while X46, X48, and X50 allow for differences in the slopes. Income is recorded in the Census from $1 to $9,999 by $10 intervals and from $10,000 to $24,999 by $1,000 intervals, with one category for $25,000 or more. The mid-points are used as the values of these intervals. For example: $0-$9 = $4.50, $10-$19 = $14.50, $20-$29 = $24.50, etc. Negative income is assumed to be a temporary phenomenon and is given the value of $4.50. Income over $7,000 is assumed to be $7,000. The abbreviated regression models revealed relationships which appear to be linear in logs to about the $7,000 income level and horizontal thereafter for most of the models esti- mated [36, pp. 55, 61, 62]. Dependency Ratio The dependency ratio is described by a binary and a continuous variable. 106 X51 = the number of persons in the household who are less than 15 and over 64 years of age, divided by the number who are 15 through 64 years of age 0 if there is no one in the household who is 15 through 64 years of age X52 = 1 if there is no one in the household who is 15 through 64 years of age 0 otherwise A linear specification was chosen after plotting some of the relationships from the contingency tables of Chapter III. The relationships exhibited considerable variation. Thus binary vari- ables would probably have described the relationships more accurately but used up valuable computer time. It was assumed that some of the variation would be removed in the multiple regression analysis and that a linear specification would be adequate. The thirteen sets of variables just described constitute the common group of independent variables that are used throughout the remainder of this study. They are used as independent variables with a series of binary regressands that are discussed next. Then they are used in a multiple regression model with the INDEX discussed in Chapter 11. They also serve as independent variables in the twenty regression models used to test the INDEX for weight sensitivity. For further information regarding these variables, see the technical documentation of the Census sample used here [36]. Endogenous Variables These next variables, presented in Table IV-l, are the ten binary regressands to be used in ten multiple regression models. A discussion of the measures of housing from which these variables are 107 TABLE IV—l.-—Binary Dependent Variables From Selected Measures of Housing Conditions Housing Condition Measures Binary Dependent Variables Number of Rooms Structural Condition Water Supply Access to a Bath or Shower Year Built Number of Bathrooms Type of Heating Equipment Access to Kitchen Facilities Access to a Telephone Y1 =1 0 Y2 =1 0 Y3 =1 0 Y4 = 0 Y5 = 0 Y6 =1 0 Y7 =1 0 Y8 =1 0 Y9 =1 0 Y10 =1 0 if the housing unit has six or more rooms. otherwise if the housing unit is struc- turally sound otherwise if the housing unit is not structurally dilapidated otherwise if hot and cold water is piped inside the housing unit otherwise if the housing unit provides ex- clusive access to a bath or shower otherwise if the housing unit was built from 1950 to 1960 otherwise if the housing unit has one or more bathrooms otherwise if the housing unit possesses the four preferred types of heating equipment: Built-in Electric Steam or Hot Water Warm Air Furnace Floor, Wall or Pipeless Furnace otherwise if the housing unit provides ex- clusive direct access to kitchen facilities otherwise if the housing unit provides access to a telephone otherwise 108 taken was provided in Chapter II. The variables Y1, Y2, ..., Y10 represent the highest level or levels of housing condition for each measure of housing condition. Assumptions and Interpretation of the Models The use of a binary dependent variable calls for a special interpretation of the models and results in violation of some of the classical assumptions of multiple regression. The special interpre- tation involves viewing the estimated regression coefficients as contributing to or detracting from the probability that the event described by the dependent variable occurs. Thus a negative coef- ficient reduces the probability that an event occurs while a positive coefficient increases that probability. This interpretation causes a problem when the prediction for an observation is less than zero or exceeds unity. The problem is approached by defining all predictions greater than unity as equal to unity and all predictions less than zero as equal to zero [10, pp. 425—428]. The classical assumptions violated here are the assumptions of homoskedasticity and normality of the error term. For a dis- cussion of the assumptions of this type of model and consequences of these assumptions, see Appendix III. Briefly this results in in- efficient and asymptotically inefficient ordinary least squares esti- mates (OLS) of the regression coefficients. However, these esti- mates are unbiased and consistent. This means that the OLS esti- mates of the variances of these coefficients are biased. The direction of this bias was not determined so the OLS estimates of the variances are not presented and no statistical tests are performed. 109 Empirical Results The ten models used to estimate these net relationships are presented in the same order that their dependent variables are pre- sented in Table IV-l. The sets of socio-economic and locational characteristics with the strongest relationships to the dependent variable in question are discussed while some of the sets are left for the reader to examine. The strength of the relationship is judged by two measures: (1) the size of the estimated parameter coupled with the range of the independent variable which is referred to as the potential effect, and (2) the relative size of the R2 delete. The R2 delete for a particular variable is the R2 for the model with that explanatory variable removed. If there were no multicollinearity between the independent variables in the model, the R2 delete would be a good indicator of the importance of the individual variable. The difference between the total R2 and the R delete would represent the percentage of the variation in the de- pendent variable directly attributable to the omitted variable. However, with multicollinearity in the total model part of the effects of the omitted variable are attributed to the included inde- pendent variables with which it is correlated. Because the models used in this study have varying degrees of multicollinearity the R2 deletes are not completely accurate indications of the importance of the omitted variable. Y : Six Rooms or More The first model presented is used to estimate the net relationships between the socio-economic and locational characteristics 110 previously discussed and the existence of six rooms or more in the housing unit. The empirical results presented in Table IV-2 indicate that the predetermined variables explained 20.4 percent of the vari- ation in the dependent variable. Several sets of these socio-economic and locational charac- teristics appear to have a larger effect on the dependent variable than other sets: household income, the dependency ratio, the age of the household head, the education of the household head, and the type of tenure. An examination of the R2 deletes reveals that the per- centage of the dependent variable explained decreases by 4.48 when the variable designating owner occupancy is dropped from the model. The probability that the housing unit possesses six rooms or more increases by .254 if the unit is owner occupied rather than renter occupied. The probability increases by .114 when the occupants pay no cash rent rather than the more typical renter status. The cate- gory, no cash rent, is usually associated with lower levels of hous- ing condition than the renter category. The variables describing household income appear to have the largest estimated effect on the probability that the housing unit has six rooms or more. If the household consists of a mixture of families and unrelated individuals, the initial effect on the probability of occupying a larger housing unit is -.657. The estimated effect of household income for this group is a positive .924 with $7,000 income or more. The intercept for the income of households consisting of families is not as negative (—.127) as that for mixed households but the slope is also less. Both intercepts represent the difference 111 TABLE IV-2.-—Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and the Presence of Six Rooms or More . . c Regression R2 Predetermined Variables Coefficient Deletes Constant Term -.6944 Region of the United States xl Northeasta .0858 .2000 X2 North Centrald ---- ---- x3 Southa .0438 .2029 X4 Westa .0800 .2013 Size of Place xS Rural Farma a .0342 .2040 X6 Rural Nonfarm .0970 .2038 X7 Urban Territory Outside of Placesa b .1083 .2037 X8 Log. of the Size of Place (Pepulation) .0250 .2037 Location Within Urbanized Area X9 In a Central City8 d .0237 .2039 X10 In Remainder or Urbanized Area ---- ---— Age of Household Head X11 Age b/10 b .3926 .2014 X12 Age Squared /1, 000 .6309 .2024 X13 Age Cubedb /100, 000 .3384 .2029 Se ex of Household Head X1 Maled ---- ---- RX1 Female3 .0233 .2038 ac of Household Head Rx16 White With Spanish Surnamea .0444 .2039 X Whited ---- ---- 17 X18 Negroa .0160 .2039 x19 Other Racea .0317 .2040 Nati 1vity and Parentage of Household Head X20 Native With Native Parents 8 ---- ---- X21 Native With One Foreign Parent .0026 .2040 X22 Native With Foreign Parents8 .0116 .2040 X23 Foreign Born8 .0346 .2037 MetrOpolitan Resigence in 1955 X24 Same House a ---- ~--- X25 Different House Same County .0434 .2027 X26 Different County Same State3 .0281 .2038 x27 Contiguous State3 .0437 .2038 X28 Noncontiguous Statea .0447 .2036 X29 Abroad or at Sea8 .0498 .2039 112 TABLE IV-2.--Continued. . . c Regression R2 Predetermined Var1ab1es Coefficient Deletes OcCUpational Classification x30 Farmera a .1084 .2033 X31 Farm Manager .1227 .2040 x32 Farm Foremana .0058 .2040 x33 Farm Laborera a .0144 .2040 X34 Farm Service Workera .0488 .2040 X35 White Collar Worker .0714 .2034 X36 Blue Collar Workera .0072 .2040 X37 Service Worker3 .0005 .2040 X38 Laborera d .0075 .2040 X39 OCCUpation Not Reported ---- "" X40 No Occupationa .0583 .2037 Tenure x41 Owneda .2541 .1592 X42 Rentedd ---- ---- X43 No Cash Renta .1141 .2023 Educational Level of Household Head X44 Educational Level if'530.5 Years .0109 .2022 x Educational Level if >10.5 Yearsb .0059 .2009 Log. of Household Income for Households of Various Types X Unrelated Individuals (Slope)b .0098 .2040 X47 Family (Intercept)a .1274 .2036 X Family (SIOpe)b .0570 .2017 X49 Mixed (Intercept)a .6568 .2033 x Mixed (Slope)b .2403 .2027 Mp endency Ratio x51 Dependency Ratiob .0658 .1958 X52 No One 14-643 .0119 .2040 R2 = .2040 aThis variable is dichotomous equalling one if the stated condition holds, zero otherwise. b O O O O This var1able 15 continuous. cThe observation unit is the household and the variables per- tain either to the household or to the head of household. dThis variable was omitted to avoid singularity. One- -in- -a- thousand sample tapes, 20 percent sample, 1960 Censuses of Population and Housing [36]. Source: 113 from the intercept for households of unrelated individuals. These empirical results for household income indicate that for households consisting of unrelated individuals income has little relationship to the probability that they occupy housing units with six rooms or more. Households consisting of families initially have a lower probability of occupying a larger housing unit but that probability increases as income increases. Households that are mixed initially have the lowest probability of occupying a larger housing unit but show a larger positive relationship with household income. At house- hold income of $7,000 or more the probability that mixed households occupy a larger housing unit exceeds that for families which exceeds that for unrelated individuals. The dependency ratio exhibits an estimated positive relation- ship to the probability that the household occupies a housing unit with six rooms or more. The maximum value of this ratio is ten which would indicate a possible estimated increase in the probability that the household occupies a larger housing unit of .658 over a household with no one under 15 or over 64 years of age. The variables describing the age of the household head exhibit a substantial positive relationship to the probability that the house- hold occupies a larger housing unit. Between the ages of 15 and 100 the probability is estimated to increase by .543. Table V-3 in Chapter V presents the estimates of this relationship. The proba- bility first increases at a decreasing rate, goes through a point of inflection at about 60 years of age then increases at an increasing rate. This estimated net relationship indicates that as age increases people tend to live in larger homes, other variables held constant. 114 The set of variables describing the education of the household head exhibit an estimated potential increase of .212 in the proba- bility that the household occupies a larger housing unit. It was hypothesized that the educational level of the household head up to ten and a half years would have a greater positive relationship to housing condition than education beyond that point. The estimated net relationships support that hypothesis. The sets of socio-economic and locational characteristics just discussed each can have a potential effect greater than .200 on the probability that the housing unit possesses six rooms or more. These characteristics appear to be the primary explanatory variables. Three other sets of characteristics have potential effects of greater than .150: size of place, occupational classification, and region of the United States. Due to space limitations these variables are not discussed. The reader may examine Table IV-2 for the effects of these variables. Y2: Structurally Sound and Y3: Not Structurally Dilapidated These two models are discussed together because their de- pendent variables represent the highest and lowest levels of struc- tural condition. The independent variables explained 19.29 percent of the variation in Y2 and only 10.82 percent in the variation in Y as indicated in Tables IV-3 and IV-4. An examination of the R2 3 deletes indicate that four variables if omitted reduce the R2 the educational level, and X de- owned, X 51 most: X18 Negro, X pendency ratio. 41 44 115 TABLE IV-3.--Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and a Structurally Sound Housing Unit . . c Regression R2 Predeterm1ned Var1ables Coefficient Deletes Constant Term .3364 Region of the United States X1 Northeasta d .0109 .1928 X2 North Central ---- ---- x3 Southa -.0216 .1925 x4 West3 .0119 .1928 Size of Place xS Rural Farma .0655 .1928 x6 Rural Nonfarma .0886 .1926 x7 Urban Territory Outside of Placesa .1740 .1918 X3 Log. of the Size of Place (Population)b .0375 .1918 Location Within Urbanized Area X9 In a Central City3 -.0220 .1928 X10 In Remainder of Urbanized Aread ---— ---- Age of Household Head X11 Ageb /10 -.0397 .1929 x12 Age Squaredb/l, 000 .0766 .1929 x13 Age Cubedb/loo, 000 -.0392 .1929 Sex of Household Head x Maled --—- —--- x14 Female3 .0077 1929 15 ‘ Race of Household Head x16 White With Spanish Surnamea -.ll76 .1916 X Whited ---- ---- 17 x18 Negroa -.1737 .1796 x19 Other Racea -.1253 .1926 Nativity and Parentage of Household Head X 0 Native With Native Parentsd ---- —--- X21 Native With One Foreign Parenta .0209 .1928 X2 Native With Foreign Parentsa .0291 .1924 X23 Foreign Borna .0683 .1909 Metropolitan Residence in 1955 X24 Same Housed ---- ---- X25 Different House Same County3 .0238 .1923 X26 Different County Same Statea .0377 .1924 X27 Contiguous Statea .0399 .1927 X28 Noncontiguous Statea .0570 .1920 X Abroad or at Seaa .0617 .1926 29 116 TABLE IV-3.--Continued. . . c Regression R2 Predeterm1ned Var1ables Coefficient Deletes Occupational Classification X3O Farmera -.0023 .1929 x31 Farm Manager3 .0648 .1929 X32 Farm Foremana -.0508 .1929 x33 Farm Laborera -.1574 .1914 x34 Farm Service Workera -.0908 .1929 X35 White Collar Workera .0387 .1926 X36 Blue Collar Workera -.0083 .1929 X37 Service Worker3 -.0052 .1929 x38 Laborera -.0825 .1920 X39 Occupation Not Reportedd ---- ---— X4O No Occupationa -.O328 .1928 Tenure X41 Ownedad .1305 .1736 X42 Rented a —--- ---- X43 No Cash Rent -.0108 .1929 Educational Level of Household Head x44 Educational Level if 510.5 Yearsg .0220 .1811 X4S Educational Level if >10.5 Years .0011 .1928 Log. of Household Income for Households of Various Types X46 Unrelated Individuals (Slope)b .0282 .1923 X47 Family (Intercept)a -.0178 .1929 X48 Family (Slope)b .0489 .1901 x49 Mixed (Intercept)a -.2717 .1927 X Mixed (SIOpe)b .1043 .1925 Depggdency Ratio XSl Dependency Ratiob -.0440 .1870 x52 No One 14 64a -.0180 .1928 R2 = .1929 8This variable is dichotomous equalling condition holds, zero otherwise. b . . . . This variable 1s continuous. one if the stated cThe observation unit is the household and the variables pertain either to the household or to the head of household. dThis variable was omitted to avoid singularity. Source: One-in-a-thousand sample tapes, 20 percent sample, Censuses of Population and Housing [36]. 1960 117 TABLE IV-4.--Estimated Net Relationships Between Socio-economic and Locational Characteristics of the OCCUpants and a Housing Unit That is Not Structurally Dilapidated . . c Regression R2 Predeterm1ned Var1ables Coefficient Deletes Constant Term .8340 Region of the United States X1 Northeasta .0023 .1082 X2 North Centrald ---- ---- X3 Southa -.0117 .1078 X4 West8 -.0043 .1082 Size of Place xS Rural Farma .0661 .1076 X6 Rural Nonfarma .0298 .1081 X7 Urban Territory Outside of Placesa b .0509 .1079 X8 Log. of the Size of Place (Population) .0130 .1077 Location Within Urbanized Area x9 In a Central Citya d .0021 .1082 X1 In Remainder of Urbanized Area ---- ---- Age 8f Household Head x11 Ageb/IO -.0517 .1079 x12 Age Squaredb/l,000 .0910 .1080 x13 Age Cubedb/100,000 -.0468 .1081 Sex of Household Head x14 Maled ---- ---- x15 Femalea .0097 .1080 Race of Household Head x16 White With Spanish Surnamea -.0291 .1079 X Whited ---- ---- 17 a X18 Negro -.1015 .0929 x19 Other Racea —.0588 .1080 Nativity and Parentage of Household Head X20 Native With Native Parents ---- ---- X21 Native With One Foreign Parenta .0039 .1082 X Native With Foreign Parentsa .0039 .1082 X Foreign Borna .0160 .1078 Metropolitan Residence in 1955 X24 Same House -—-- ---- X25 Different House Same County8 .0119 .1077 X26 Different County Same Statea .0161 .1078 x27 Contiguous Statea .0177 .1080 X28 Noncontiguous Statea .0209 .1078 X Abroad or at Seaa .0107 .1082 29 118 TABLE IV-4.--Continued. ._ .....__—_..-—- . . c Regression R2 Predeterm1ned Var1ables Coefficient Deletes Occupational Classification x30 Farmera -.0084 .1082 X31 Farm Managera .0713 .1081 X32 Farm Foremana .0534 .1081 X33 Farm Laborera -.0871 .1066 x34 Farm Service Workera —.2234 .1080 X35 White Collar Workera .0149 .1080 X36 Blue Collar Workera .0095 .1081 X37 Service Worker3 .0107 .1081 x38 Laborera -.0277 .1078 X39 Occupation Not Reportedd ---- ---- X40 No Occupationa -.0084 .1082 Tenure x41 Owneda .0427 .1013 X42 Rented a ---- ---- X43 No Cash Rent -.0320 .1074 Educational Level of Household Head X Educational Level if 510.5 Yearsb .0096 .1006 x45 Educational Level if >10.5 Yearsb -.0004 .1081 Log. of Household Income for Households of Various Types X46 Unrelated Individuals (Slope)b .0108 .1079 X47 Family (Intercept)a -.0163 .1082 X Family (Slope) .0210 .1064 x49 Mixed (Intercept)a -.l731 .1079 X Mixed (Slepe)b .0611 .1077 Mp endency Ratio b X51 Dependency Ratio -.0200 .1041 X No One 14-64a -.0014 .1082 52 R2 = .1082 aThis variable is dichotomous equalling condition holds, zero otherwise. b . . . . This variable 1s continuous. one if the stated cThe observation unit is the household and the variables pertain either to the household or to the head of household. dThis variable was omitted to avoid singularity. Source: One-in-a—thousand sample tapes, 20 percent sample, Censuses of Population and Housing [36]. 1960 119 With both of these dependent variables the white household head is associated with the highest level of housing condition, white with a Spanish surname next, followed by other race, and Negro last. The Negro household head has an estimated .174 lower probability of occupying a structurally sound housing unit than a white household head. He has an estimated .102 lower probability of occupying a housing unit that is not structurally dilapidated. The owner occupied housing units are more likely to possess higher levels of structural condition than renter occupied units which possess higher condition levels than units where the occupants pay no cash rent. The educational level of the household head is positively related to structural condition. A household head with ten and a half years education or more is an estimated .249 more likely to occupy sound housing than one with no education. The hypothesized decrease in the positive relationship between education of the house- hold head and housing condition at ten and a half years of education is supported by parameter estimates of the first model. The second model with not structurally dilapidated as a dependent variable exhibits a small negative relationship with housing condition after ten and a half years of education. The dependency ratio exhibits a relatively strong negative relationship with structural condition. As the dependency ratio varies over its observable range from 0 to 10, the probability that the housing unit is structurally sound decreases by .440 and the probability that it is not dilapidated increases by .200. These 120 estimated net relationships indicate that the higher the proportion of household members under 15 and over 64 the lower the structural condition of the unit. Three other sets of characteristics appear to have relatively strong relationships to structural condition: age, household income, and occupational classification. The different occupational classifi- cations exhibited an estimated .222 probability range between the classification where the household is most likely and the classifi- cation where the household is least likely to occupy a structurally sound housing unit. Listed from the occupational classification where the household is most likely to the one where the household is least likely to occupy sound housing, the classifications are arranged in this order: farm manager, white collar worker, no occupation reported, farmer, service worker, blue collar worker, no occupation, farm foreman, laborers, farm service workers, and farm laborers. The relationships between occupational classifications and structural condition are different but follow the same basic pattern for the second model, the model with not structurally dilapidated as dependent variable. The occupational classifications exhibit an estimated .295 range in the probability that the household does not occupy dilapi- dated housing. The household income of various types of households exhibit a similar pattern of relationships to structurally sound and structurally dilapidated as was exhibited with the dependent variable, six rooms or more. Households composed of unrelated individuals exhibit a positive relationship between household income and structural 121 condition. This relationship is smaller than that for households composed of families and smaller yet than the relationship for house— holds composed of families and unrelated individuals. However, initially households composed of unrelated individuals have a higher probability of occupying housing units with higher levels of struc- tural condition than households composed of families. These family type households in turn have a higher probability of occupying struc- turally desirable housing than mixed households. The last set of socio-economic and locational variables to be discussed here, age of the household head, exhibits an unusual re— lationship to structural condition. The probability that the house— hold occupies structurally sound housing first decreases at a decreas- ing rate, reaching a minimum at approximately 35 years of age, then increases first at an increasing rate and later at a decreasing rate reaching a peak at about 95 years of age. These estimated relation- ships have been calculated and are presented in Table V-3 of Chapter V. The relationships between age of the household head and struc- turally dilapidated are similar to those for structurally sound with maximums and minimums occurring at different age levels. Tables IV-3 and IV-4 may be examined to determine the direction and magnitude of relationships between other socio-economic and locational characteristics and structural conditions. Y4: Hot and Cold Water Piped Inside the Housing Unit Five sets of socio-economic and locational characteristics appear to have the predominant effects on the presence of hot and 122 cold water piped into the housing unit: household income, occu- pational classifications, education of the household head, race, the region of the United States, and the size of place. The empirical results from this model are presented in Table IV-S. The household income variables exhibit a pattern of relation- ships with the water supply measure of housing condition which is the same as the pattern exhibited with the number of rooms and structural condition. This set of characteristics is not discussed other than to note the range of effects of household income for the various types of households. As income goes from 0 to $7,000 or more, the effect on the probability that the housing unit has hot and cold water piped inside goes from O to .096 for households composed of unrelated individuals, from -.068 to .152 for households composed of families, and from -.398 to .165 for mixed households. The occupational classifications exhibit a .308 probability range between the classification where the household has the highest probability of having the desirable water supply conditions and the classification where the household has the lowest probability. Listed from the highest to the lowest probability the classifications relate to water supply in this order: farm manager, farm foreman, white collar worker, service worker, occupation not reported, blue collar worker, no occupation, farmer, laborers, farm service worker, farm laborer. The educational level of the household head exhibits a posi- tive relationship to hot and cold water being piped inside the housing unit between 0 and ten and a half years' education. After that point 123 TABLE IV—5.--Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and Hot and Cold Water Piped Inside the Housing Unit . . c Regression R2 Predeterm1ned Var1ables Coefficient Deletes Constant Term .4131 Region of the United States X1 Northeasta d .0260 .3055 X2 North Central ---- ---- x3 Southa -.0660 .3004 X4 Westa .0254 .3057 Size of Place xS Rural Farm3 -.0372 .3063 X6 Rural Nonfarma -.0235 .3063 X7 Urban Territory Outside of Places3 .1126 .3056 x8 Log. of the Size of Place (Population)b .0277 .3055 Location Within Urbanized Area X9 In a Central City3 .0136 .3063 X10 In Remainder of Urbanized Aread ---- ---- Age of Household Head x11 Ageb/lo .0262 .3063 x12 Age Squaredb/1,000 -.0477 .3063 x13 Age Cubedb/100,000 .0336 .3063 Sex of Household Head x14 Maled ---- ---- X15 Femalea .0163 .3061 Race of Household Head X16 White With Spanish Surnamea -.0533 .3060 X17 Whited ---- ---- X18 Negroa -.l708 .2875 x19 Other Race3 -.0808 .3062 Nativity and Parentage of Household Head X2 Native With Native Parentsd ---— ---- X21 Native With One Foreign Parenta .0060 .3063 X22 Native With Foreign Parentsa -.0112 .3062 X23 Foreign Borna .0378 .3054 MetrOpolitan Residence in 1955 X24 Same Housed ---- ---- X25 Different House Same Countya .0263 .3052 X26 Different County Same Statea .0441 .3052 X27 Contiguous Statea .0444 .3059 X28 Noncontiguous Statea .0612 .3047 X29 Abroad or at Seaa .0551 .3060 124 TABLE IV—S.--Continued . . c Regression R2 Predetermined Var1ables Coefficient Deletes Occupational Classification X30 Farmera -.0503 .3060 x31 Farm Managera .1397 .3063 x32 Farm Foremana .0442 .3063 X33 Farm Laborera -.l680 .3038 X34 Farm Service Worker3 -.1074 .3063 X35 White Collar Workera .0093 .3063 X36 Blue Collar Workera -.0009 .3064 x37 Service Workera .0013 .3064 X38 Laborera -.0716 .3054 X39 Occupation Not Reportedd —--- ---- X40 No Occupationa -.0400 .3060 Tenure X41 Owneda .0536 .3016 X42 Rentedd ---- ---- X43 No Cash Renta -.0901 .3038 Educational Level of Household Head X Educational Level if 510.5 Yearsb .0240 .2858 X Educational Level if >10.5 Yearsb -.0008 .3062 Log. of Household Income for Households of Various Types x Unrelated Individuals (Slope)b .0249 .3057 X47 Family (Intercept)a -.O682 .3061 x48 Family (Slope)b .0573 .3007 X49 Mixed (Intercept)a -.3977 .3057 x50 Mixed (Slope)b .1463 .3052 Dependency Ratio XSl Dependency Ratiob -.0199 .3046 X$2 No One 14-64a -.0027 .3064 R2 = .3064 aThis variable is dichotomous equalling condition holds, zero otherwise. b . . . . This variable is continuous. one if the stated cThe observation unit is the household and the variables pertain either to the household or to the head of household. d . . . . . . This variable was omitted to av01d Singularity. Source: One-in-a-thousand sample tapes, 20 percent sample, 1960 Censuses of Population and Housing [36]. 125 the relationship is negative but small. The probability that the housing unit has hot and cold water piped inside increases by .252 as the educational level increases from 0 to ten and a half years. The race variables account for an estimated .171 change in the probability that the housing unit contains hot and cold water piped inside. The R2 deletes indicate that the individual variable X18, Negro, if omitted would reduce R2 by almost as much as X44, education. The Negro household head has the lowest estimated proba- bility of occupying a housing unit with the desirable water supply; the other race household head has the next higher; then the household head who is white with a Spanish surname; and the white household head has the highest estimated probability of occupying a housing unit with the desirable type of water supply. The region of the United States variables account for only .092 change in the probability of the desirable water supply. How- ever, the R2 deletes indicate that omitting the variable designating the South would reduce R2 by more than is indicated for most of the other variables. The Northeast and the West have the highest esti- mated probability of having the desirable water supply. The North Central region exhibited a lower estimated probability and the South exhibited the lowest probability of a housing unit having hot and cold water piped inside. The size of place variables account for an estimated .203 change in the probability of a unit having hot and cold water piped inside. Rural farm has the lowest probability and rural nonfarm next. Then as the size of place increases the change in the 126 probability that a housing unit has the desirable water supply increases from .166. The residence category, urban territory outside of places shows an increase in this probability over the residence categories of rural farm and rural nonfarm. Relationships between other sets of socio-economic and locational characteristics and water supply can be observed in Table IV-S. Y : Exclusive Access to Bath or Shower The next model presented has exclusive access to a bath or shower as dependent variable. Four sets of socio-economic and locational characteristics exhibit substantial estimated net relation- ships with this binary variable: household income, occupational classification, educational level, and the size of place. Household income exhibits relationships similar to those exhibited with other measures of housing condition previously dis- cussed. That is income is positively related to the probability that the household has exclusive access to a bath or shower. The slope of this log-linear relationship is greatest for mixed house- holds, less for households composed of families, and the smallest for households composed of unrelated individuals. These relationships differ from the relationships previously discussed. With zero or negative income the probability of having exclusive access to a bath or shower is greater for households composed of families than for households composed of unrelated individuals. This has not been the case with previous models. 127 The occupational classification of farm manager exhibits a probability of having exclusive access to a bath or shower which is .366 greater than that for farm laborers. The pattern of relation- ships is similar to the estimated net relationships with the proba- bility that a housing unit has hot and cold water piped inside and the probability that a housing unit has one more bathroom. The educational level of the household accounts for an esti- mated .253 increase in the probability that a household has exclusive access to a bath or shower from the zero educational level to ten and a half years of education. After that point the relationship is slightly negative, decreasing .010 for each additional year of edu- cation beyond ten and a half years. The R2 delete indicates that if X44 were dropped from the model the percent of the variation in the dependent variable explained would decrease by 1.83. The size of place variables can account for an estimated .214 change in the probability that a household has exclusive access to a bath or shower. As could be expected the rural farm and rural non— farm categories exhibit the lowest estimated probability of possessing this desirable housing characteristic. Urban territory outside of places has a higher estimated probability. The logarithm of the size of place has a positive relationship and increases the estimated probability of exclusive access to a bath or shower by .124 for places of one million or more population. These sets of characteristics--household income, occupational classifications, educational level, and size of place-~exhibit the strongest estimated net relationships with the dependent variable. 128 However, two other individual variables indicate by their R2 deletes that they explain a substantial pr0portion of the observed variation in the dependent variable. They are X Negro and X14, owned. Other 18’ relationships can be observed in Table IV-6. Y : Built from 1950 to 1960 The dependent variable for this next model indicates whether the housing unit was built within the decade previous to the Census. As indicated in Table IV-7, 23.1 percent of the variation in this regressand is explained by the independent variables used. A differ- ent mix of regressors appear to be the primary explanatory variables in this model than in the previous models: metropolitan residence in 1955, tenure, and age of the household head. The metropolitan residence in 1955 variables account for an estimated .311 change in the probability that the housing unit was built from 1950 to 1960. However, the relationship is somewhat irregular. As one moves from the variable indicating no move through the variables indicating moves of increasing distance, the probability does not increase smoothly. It increases from X same house, to 24’ X different county same state; decreases to X contiguous state; 26’ 27 reaches a maximum at X28, noncontiguous state; and decreases to X29, abroad or at sea. The tenure variables can account for an estimated .248 change in the probability that the housing unit was built from 1950 to 1960. According to the R2 deletes, if the owned tenure category, X 1, were 4 omitted from the model the percentage of the dependent variable explained would decrease by 4.96. The rented tenure category, X42 129 TABLE IV-6.--Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and Ex- clusive Access to a Bath or Shower . 2 . . c Regression R Predeterm1ned Var1ables Coefficient Deletes Constant Term .3038 Region of the United States x Northeasta .0336 .2720 X2 North Centrald ---- ---- X3 Southa —.0323 .2720 X4 West3 .0363 .2722 Size of Place X5 Rural Farm8 —.0899 .2729 x Rural Nonfarma -.0542 .2732 x7 Urban Territory Outside of Placesa b .0882 .2729 X3 Log. of the Size of Place (Population) .0206 .2729 Location Within Urbanized Area x9 In a Central Citya d .0079 .2733 X10 In Remainder of Urbanized Area ---- ---- Age of Household Head x11 Ageb/IO —.0038 .2733 x12 Age Squaredb/1,000 .0269 .2733 x13 Age Cubedb/IOO,OOO -.0206 .2733 Sex of Household Head X14 Maled ---- ---- X15 Femalea .0535 .2711 Race of Household Head x16 White With Spanish Surnamea -.0410 .2731 - d X17 White ---- ---- X18 Negroa -.l440 .2616 X19 Other Racea -.0454 .2732 Nativity and Parentage of Household Head X20 Native With Native Parentsd ---- ---- X21 Native With One Foreign Parenta .0143 .2732 X Native With Foreign Parentsa .0003 .2733 X23 Foreign Borna .0545 .2716 Metropolitan Residence in 1955 X24 Same Housed ---- -—-- X25 Different House Same County3 .0228 .2726 X26 Different County Same State8 .0396 .2725 X27 Contiguous State8 .0371 .2730 x28 Noncontiguous Statea .0402 .2727 X29 Abroad or at Sea8 .0279 .2732 130 TABLE IV-6.--Continued _____._.. _—__—c——. - o—o—n— —-—.— .__..—_-- . . c Regression R2 Predetermined Var1ables Coefficient Deletes Occupational Classification X30 Farmera -.0411 .2731 x31 Farm Manager3 .2210 .2731 x32 Farm Foreman: .0719 .2733 X33 Farm Laborer —.l453 .2716 X34 Farm Service Workera -.0517 .2733 X35 White Collar Workera .0257 .2731 x36 Blue Collar Workera .0143 .2732 x37 Service Worker3 -.0113 .2733 x38 Laborera d -.0617 .2727 X39 Occupation Not Reported ---- ---- X40 No Occupationa —.Ol79 .2732 Tenure x Owneda .0781 .2644 X41 Rentedd ---- ---- 42 X43 No Cash Renta -.O751 .2717 Educational Level of Household Head X Educational Level if 510.5 Years .0241 .2550 X45 Educational Level if >10.5 Yearsb -.0001 .2733 Log. of Household Income for Households of Various Types x46 Unrelated Individuals (Slope)b .0303 .2724 X47 Family (Intercept)a .0118 .2733 x48 Family (Slope)b .0659 .2668 x49 Mixed (Intercept)a -.2923 .2730 x50 Mixed (Slope)b .1462 .2723 Dependency Ratio b X51 Dependency Ratio -.0105 .2729 x52 No One 14-64a .0191 .2732 R2 = .2733 aThis variable is dichotomous equalling condition holds, zero otherwise. b . . . . This variable is continuous. one if the stated cThe observation unit is the household and the variables pertain either to the household or to the head of household. dThis variable was omitted to avoid singularity. Source: One-in-a-thousand sample tapes, 20 percent sample, 1960 Censuses of P0pulation and Housing [36]. 131 TABLE IV-7.--Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and the Housing Unit Being Built From 1950 to 1960 ..— . . c Regression R2 Predetermined Var1ables Coefficient Deletes Constant Term -.4905 Region of the United States X1 Northeasta -.0129 .2311 X2 North Centrald ---- ---- X3 Southa .0935 .2254 X4 Westa .0863 .2276 Size of Place XS Rural Farm3 -.0273 .2312 x6 Rural Nonfarma .0779 .2310 X Urban Territory Outside of Placesa b .2117 .2300 X8 Log. of the Size of Place (Population) .0139 .2311 Location Within Urbanized Area X9 In a Central City8 -.0734 .2299 X10 In Remainder of Urbanized Aread ---- ---- Age of Household Head x11 Ageb/lo .2987 .2294 x12 Age Squaredb/1,000 -.6768 .2290 x13 Age Cubedb/100,000 .4210 .2292 Sex of Household Head x14 Maled X15 Femalea -.0325 .2308 Race of Household Head X16 White With Spanish Surnamea -.0411 .2311 X1 Whited ---- ---— X18 Negroa -.0050 .2312 X19 Other Racea .0061 .2312 Nativity and Parentage of Household Head X20 Native With Native Parentsd -—-- ---— X2 Native With One Foreign Parenta .0155 .2311 X22 Native With Foreign Parentsa .0449 .2302 X23 Foreign Borna .0506 .2304 MetrOpolitan Residence in 1955 X24 Same Housed ---- ---- X25 Different House Same Countya .2186 .1930 x26 Different County Same Statea .2602 .2120 X27 Contiguous Statea .2552 .2234 X28 Noncontiguous State8 .3109 .2106 X Abroad or at Seaa .2681 .2269 29 132 TABLE IV-7.--Continued . . c Regression R2 Predetermined Variables Coefficient Deletes Occupational Classification X30 Farmera -.0612 .2309 X31 Farm Managera -.1075 .2312 X32 Farm Foremana -.0079 .2312 X33 Farm Laborera -.0981 .2308 X34 Farm Service Worker3 -.0914 .2312 X35 White Collar Workera .0199 .2312 X36 Blue Collar Workera -.0163 .2312 x37 Service Worker3 —.0206 .2312 X38 Laborera -.0610 .2309 X39 OCCUpation Not Reportedd ---- ---- X40 No OCCUpationa -.0115 .2312 Tenure x41 Owneda .2477 .1816 X42 Rentedd ---- ---- X43 No Cash Renta .0604 .2306 Educational Level of Household Head x44 Educational Level if 510.5 Yearsb .0035 .2310 x45 Educational Level if >10.s Yearsb .0037 .2298 Log. of Household Income for Households of Various Types x46 Unrelated Individuals (Slope)b .0032 .2312 X47 Family (Intercept)a -.0777 .2310 x48 Family (Slope)b .0319 .2304 X49 Mixed (Intercept)a -.0832 .2312 Mixed (Slope)b .0155 .2312 Dependency Ratio x51 Dependency Ratiob .0063 .2311 X52 No One 14-64a .0454 .2307 R2 = .2312 aThis variable is dichotomous equalling one if the stated condition holds, zero otherwise. b . . . . This variable is continuous. cThe observation unit is the household and the variables pertain either to the household or to the head of household. dThis variable was omitted to avoid singularity. Source: One-in-a-thousand sample tapes, 20 percent sample, 1960 Censuses of POpulation and Housing [36]. 133 exhibits the lowest estimated probability that the housing unit was built during the decade prior to the Census. The tenure category, no cash rent, has a higher probability of having this desirable hous- ing characteristic and the tenure categorysowned, exhibits the highest probability. The variables describing the age of the household head can account for an estimated .220 change in the probability that the occupied housing unit was built from 1950 to 1960. The relationship which is also presented in Table V-3 increases to a maximum at 31 years of age, decreases to a minimum at 76 years of age, and then increases. The maximum and minimum are specified within the range 15 to 95 years of age. The first portion of this relationship seems plausible. The last portion which turns up, however, appears suspect. This relationship will be discussed further in Chapter V where the relationships between the age of the household head and all Of the dependent variables are considered. With this dependent variable the R2 deletes point to these same sets of characteristics as the primary explanatory variables. The estimated relationships between other sets of socio-economic and locational characteristics and the probability that the housing unit was built from 1950 to 1960 may be examined in Table IV-7. Y : One or More Bathrooms The dependent variable for this next model records the presence of one or more bathrooms in the housing unit. As Table IV-8 indicates the independent variables explain 29.3 percent of the variation in this 134 TABLE IV-8.--Estimated Net Relationships Between Socio-economic and Locational Characteristics and One or More Bathrooms in the Housing Unit . . c Regression R2 Predeterm1ned Variables Coefficient Deletes Constant Term .2142 Region of the United States Xl Northeasta .0336 .2921 X2 North Central ---- ---- X Southa .0425 .2913 x4 Westa .0429 .2918 Size of Place x Rural Farma .0594 .2931 x6 Rural Nonfarma .0183 .2932 X Urban Territory Outside of Placesa .1312 .2925 X8 Log. of the Size of Place (Population)b .0308 .2924 Location Within Urbanized Area X9 In a Central Citya .0010 .2932 X10 In Remainder of Urbanized Aread ---- ---- Age of Household Head x11 Ageb/lo .0037 .2932 x12 Age Squaredb/l,0-0 .0268 .2932 x13 Age Cubedb/100,000 .0202 .2932 Sex of Household Head X14 Maled --—— ---- x1 Femalea .0479 .2917 Race of Household Head X White With Spanish Surnamea .0642 .2928 xi? mad X Negroa .1684 .2790 X19 Other Race3 .0914 .2930 Nativity and Parentage of Household Head X20 Native With Native Parentsd ---- ---- X2 Native With One Foreign Parenta .0172 .2931 X22 Native With Foreign Parentsa .0011 .2932 X23 Foreign Borna .0556 .2917 Metropolitan Residence in 1955 X24 Same House ---— ---- X25 Different House Same Countya .0287 .2922 X26 Different County Same Statea .0483 .2922 X27 Contiguous Statea .0486 .2928 X28 Noncontiguous Statea .0534 .2923 x Abroad or at Seaa .0390 .2931 29 135 TABLE IV.8.--Continued Predetermined Variablesc Eggffiziggt De1:tes OccuPational Classification x30 Farmera .0507 .2930 x31 Farm Managera .1926 .2931 x32 Farm Foremana .0559 .2932 X33 Farm Laborera .1461 .2917 X34 Farm Service Workera .0369 .2932 x35 White Collar Workera .0293 .2931 x36 Blue Collar Workera .0091 .2932 x37 Service Workera .0141 .2932 x38 Laborera d .0730 .2924 X39 Occupation Not Reported ---- ---- x40 No Occupationa .0256 .2931 Tenure x41 Ownedad .0889 .2830 X42 Rented ---- ---- x43 No Cash Renta .0607 .2923 Educational Level of Household Head X44 Educational Level if 510.5 Years .0261 .2743 X45 Educational Level if >10.5 Years .0000 .2932 Log. of Household Income for Households of Various Types x46 Unrelated Individuals (Slope)b .0321 .2923 X47 Family (Intercgpt)a .0031 .2932 X48 Family (Slope) .0710 .2865 x49 Mixed (Intercgpt)a .3269 .2929 XSO Mixed (SlOpe) .1572 .2922 Dependency Ratio b X$1 Dependency Ratio .0137 .2926 x52 No One 14.64a .0178 .2931 R2 = .2932 aThis variable is dichotomous equalling one if the state condition holds, zero otherwise. b . . . . This variable is continuous. cThe observation unit is the household and the variables pertain either to the household or to the head of household. dThis variable was omitted to avoid singularity. Source: One-in-a-thousand sample tapes, 20 percent sample, 1960 Censuses of Population and Housing [36]. d 136 dependent variable. The sets of socio-economic and locational characteristics which have the strongest estimated relationships with the regressand are: size of place, occupational classification, edu- cation of the household head, and household income. The size of place variables exhibit an estimated .244 net effect on the probability that the housing unit contains one or more bathrooms. The rural farm residence category has the lowest proba- bility and the rural nonfarm residence category has .041 greater probability for possessing the desirable housing characteristics. Housing units in urban territories outside of places have a .191 greater probability of possessing the desirable housing character- istics than units in the rural farm residence category. The logarithm of the size of place exhibits a positive relationship with the regressand. Housing units in places of one million and more population have a .244 greater probability of containing one or more bathrooms than units in the rural farm residence category. The occupational classifications can explain an estimated .339 change in the probability that a housing unit contains one or more bathrooms. The farm manager classification exhibits the highest probability and the farm laborer, the lowest. The occupational classifications exhibit almost the same relationships with this dependent variable as they exhibit with two other dependent vari- ables: Y4, hot and cold water piped inside and Y5, exclusive access to a bath or shower. Notice that both the farm manager and farm foreman classifications exhibit a higher probability of one or more bathrooms than the white collar worker classification. 137 The education of the household head shows a familiar estimated net relationship to this measure of housing condition. The estimated probability that the occupied housing unit contains one or more bath- rooms increases from zero to .274 as the education of the household head goes from zero to ten and a half years. Beyond that amount of education the estimated relationship is zero. The R2 deletes indi- cate that drOpping this first education variable from the model would reduce the total R2 by .019. The household income variables also exhibit a familiar relationship with this measure of housing condition. The logarithm Of income exhibits a positive relationship with the probability that the housing unit contains one or more bathrooms. The logarithm of household income for mixed households showed the largest estimated relationship with the regressand, households consisting of families next, and households consisting of unrelated individuals showed the smallest relationship. With zero or negative income households con- sisting of families had the highest probability of enjoying the desirable housing characteristics, unrelated individuals next, and mixed households last. An examination of the R2 deletes reveals two variables not discussed above which appear to be important: X Negro and X 18’ 41’ owned. Households with Negro household heads have a .188 lower probability of occupying housing with one or more bathrooms than households with white household heads. Also owner occupied housing has a .150 higher probability of having this desirable housing characteristic than housing occupied by tenants who pay no cash rent. 138 These and other estimated net relationships may be observed in Table IV—8. Y : Heating Equipment The dependent variable in this model indicates the presence of one of four types of heating equipment: (1) built-in electric, (2) steam or hot water, (3) warm air furnace, or (4) floor, wall, or pipeless furnace. Table IV-9 indicates that the independent variables explained 33.4 percent of the total variation in this dependent variable. This regression model has the highest R2 of the ten multiple regression models used in this study that have bi- nary dependent variables. Five sets of sociO-economic characteristics have an estimated net effect greater than .200 on the probability that the housing unit possesses the desirable types of heating equipment. They are: region of the United States, size of place, race of household head, education, and household income. The region of the United States variables exhibit an esti- mated .367 effect on the dependent variable. The South has the lowest estimated probability that housing units contain the desirable types of heating equipment. The West has an estimated .178 greater probability. Housing units in the North Central region are .129 more likely than those located in the West to have the desirable types of heating equipment. Housing units located in the Northeast have an estimated .367 greater probability than those located in the South--the region with the lowest probability. The R2 deletes indi— cate that the percentage of the dependent variable explained would drop by .058 if South, X3 were omitted from the model. 139 TABLE IV-9.--Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and Four Desirable Types of Heating Equipment - 2 Predetermined VariablesC S:§f:::1::t De1etes Constant Term -.2722 Region of the United States x1 Northeasta d .0599 .3319 X2 North Central ---- ---- x Southa -.3072 .2762 x4 Westa -.1293 .3265 Size of Place x Rural Farma .1677 .3333 x6 Rural Nonfarma .2350 .3324 X Urban Territory Outside of Places3 b .4431 .3290 X8 Log. of the Size of Place (Population) .0955 .3292 Location Within Urbanized Area X In a Central Citya -.0668 .3330 X10 In Remainder of Urbanized Aread ---- --—- Age of Household Head x11 Ageb/IO b .1614 .3335 X12 Age Squargd /l,000 -.2791 .3336 X13 Age Cubed /100,000 .1457 .3338 Sex of Household Head x14 Maled ---- ---- x15 Femalea -.0075 .3340 Race of Household Head x16 White With Spanish Surnamea —.1472 .3326 X Whited ---- -—-- l7 X18 Negroa -.0961 .3313 X19 Other Racea -.2180 .3333 Nativity and Parentage of Household Head X20 Native With Native Parentsd a ---- -—-- X21 Native With One Foreign Parent .0230 .3338 x Native With Foreign Parentsa .0097 .3339 X23 Foreign Borna .0543 .3331 MetrOpolitan Residence in 1955 X24 Same House a ---- ---- XZS Different House Same Countya .0400 .3328 X26 Different CountyaSame State .0541 .3332 X27 Contiguous State .0784 .3333 X28 Noncontiguous State .0506 .3335 x Abroad or at Sea8 .0872 .3336 29 140 TABLE IV-9.--Continued Regression R . . c Predetermined Variables Coefficient Del 2 etes Occupational Classification x30 Farmera -.0385 X31 Farm Managera .0225 X32 Farm Foremana -.0234 X33 Farm Laborera -.1ll7 X34 Farm Service Workera .0836 x35 White Collar Workera .0468 X36 Blue Collar Workera -.Ol88 X37 Service Worker8 -.0325 X38 Laborera d -.1032 X39 Occupation Not Reported ---- X No OCCUpationa -.0142 40 Tenure x41 Owneda .1415 X42 Rentedd ---- X43 No Cash Renta .0280 Educational Level of Household Head x44 Educational Level if 510.5 Yearsb .0172 X45 Educational Level if >10.5 Yearsb .0057 Log. of Household Income for Households of Various Types X46 Unrelated Individuals (SlOpe)b .0275 X47 Family (Intercept)a -.0981 X48 Family (Slope)b .0643 X49 Mixed (Intercgpt)a -.2877 X50 Mixed (SlOpe) .1106 Dependency Ratio x51 Dependency Ratiob -.0125 x52 No One 14 64a .0103 R2 = .3304 .3339 .3340 .3340 .3335 .3340 .3337 .3339 .3339 .3331 .3340 .3191 .3339 .3292 .3308 .3336 .3337 .3308 .3338 .3337 .3337 .3340 aThis variable is dichotomous equalling one if the state condition holds, zero otherwise. b . . . . This variable is continuous. cThe observation unit is the household and the variables pertain either to the household or to the head of household. dThis variable was omitted to avoid singularity. Source: One-in-a-thousand sample tapes, 20 percent sample, 1960 Censuses of Population and Housing [36]. d 141 The size of place exhibits an estimated .405 effect on the probability that a housing unit has the desirable types of heating equipment. The rural farm and rural nonfarm residence categories have the lowest probability. The logarithm of the size of place is positively related to the probability that housing units have the desirable types of heating equipment. Housing units located in places of one million or more have an estimated .405 greater proba- bility than those in the rural farm residence category. The housing units located in urban territories outside of places have an esti- mated .275 greater probability of possessing the desirable types of heating equipment than those located in the rural farm residence category. The race Of the household head variables account for an estimated .218 effect on the probability that a housing unit has the desirable types of heating equipment. The housing units occupied by white household heads have the highest probability followed by these race categories listed from the highest probability to the lowest: Negro, white with a Spanish surname, and other race. The race cate- gory, Negro household head, results in the lowest probability that the housing unit has desirable housing characteristics for most measures of housing condition. However, with this measure, Negro has the next to highest probability for the housing units containing the desirable types of heating equipment. The education of the household head exhibits an estimated .275 effect on the probability that the housing unit has the de- sirable types of heating equipment. The effect on the probability 142 ranges from zero with no education to .275 with five or more years of college. The relationship between zero and ten and a half years of education is greater than for education beyond that point. The household income for various types of households exhibits a pattern of estimated net relationships which has been observed be- fore. The logarithm of household income for mixed households exhibits the strongest positive relationship with the probability that the housing unit has the desirable type of heating equipment, followed by households composed of families and then households of unrelated individuals. At zero levels Of household income, mixed households exhibit the lowest, households composed of families next, and house- holds of unrelated individuals the highest estimated probability of enjoying this desirable housing characteristic. An examination of the R2 deletes reveals that one other variable, not discussed above, has a substantial effect on the per- centage of the dependent variable explained. If owned, X were 41 omitted from the model the total R2 would decrease by .015. The estimated net relationships between these and other variables and the probability that a housing unit contains the desirable types of heating equipment may be observed in Table IV-9. Y : Exclusive Access to Kitchen Facilities The dependent variable for this next model indicates if the housing unit provides exclusive access to kitchen facilities. As indicated in Table IV-lO, only 7.12 percent of the total variation in this dependent variable was explained by the predetermined 143 TABLE IV-lO.--Estimated Net Relationships Between Socio—economic and Locational Characteristics of the Occupants and Ex- clusive Access to Kitchen Facilities - 2 Predetermined VariablesC g:g;g::::2t Defietes Constant Term .9273 Region of the United States x1 Northeasta .0007 .0712 X2 North Centrald ---- ---- X3 Southa .0038 .0711 x4 Westa .0011 .0712 Size of Place xS Rural Farm8 a -.0141 .0711 X6 Rural Nonfarm -.0118 .0712 X7 Urban Territory Outside of Places8 -.0133 .0712 X8 Log. of the Size of Place (Population)b -.0035 .0711 Location Within Urbanized Area x9 In a Central Citya .0013 .0712 X10 In Remainder of Urbanized Aread ---- ---- Age of Household Head x11 Ageb/IO b -.Ol85 .0711 X12 Age Squared /l,000 .0429 .0711 x13 Age Cubedb/100,OOO -.0307 .0710 Sex of Household Head X14 Maled ---- ---- x15 Femalea .0417 .0592 Race of Household Head x16 White With Spanish Surnamea -.0030 .0712 x17 Whited ---- ---- X13 Negroa -.0068 .0710 x19 Other Race3 -.0390 .0709 Nativity and Parentage of Household Head X20 Native With Native Parentsd ---- ---- x21 Native With One Foreign Parenta -.0003 .0712 x Native With Foreign Parents3 -.0023 .0712 X23 Foreign Born3 -.0004 .0712 Metropolitan Resigence in 1955 X24 Same House a ---- ---- X25 Different House Same Countya -.0020 .0712 X26 Different CountyaSame State -.0038 .0712 X27 Contiguous State a -.0119 .0710 X28 Noncontiguous State -.0082 .0710 x Abroad or at Sea8 -.0043 .0712 29 144 TABLE IV-10.--Continued . . c Regression R2 Predetermined Variables CCOefficient Deletes Occupational Classification X30 Farmera .0147 .0710 x31 Farm Manager3 .0195 .0712 X32 Farm Foreman: .0208 .0712 X33 Farm Laborer a .0156 .0710 X34 Farm Service Workera .0372 .0712 X35 White Collar Worker .0095 .0710 x36 Blue Collar Workera .0106 .0710 X37 Service Worker3 -.0095 .0711 X38 Laborera d .0057 .0712 X39 OCCUpation Not Reported ---- ---- X40 No Occupationa .0077 .0711 Tenure X41 Ownedad .0169 .0675 X42 Rented ---- ---- x43 No Cash Renta .0043 .0712 Educational Level of Household Head x44 Educational Level if 510.5 Yearsb -.0001 .0712 x45 Educational Level if >10 5 Yearsb .0001 .0712 Log. of Household Income for Households of Various Types X46 Unrelated Individuals (Slope)b .0007 .0712 X47 Family (Intercept) .0654 .0691 x48 Family (Slope)b .0060 .0707 X49 Mixed (Intercept)a .0469 .0712 X50 Mixed (SlOpe)b .0087 .0712 De epe ndency Ratio X51 Dependency Ratiob .0011 .0712 x52 No One 14 64a .0132 .0706 R2 = .0712 aThis variable is dichotomous equalling one if the stated condition holds, zero otherwise. b . . . . This variable is continuous. cThe Observation unit is the household and the variables pertain either to the household or to the head of household. dThis variable was omitted to avoid singularity. Source: One—in-a-thousand sample tapes, 20 percent sample, 1960 Censuses of Population and Housing [36]. 14S variables. Four sets Of socio-economic and locational characteristics exhibit effects which exceed .040 on the probability that the housing unit contains this desirable housing characteristic. They are: age of the household head, sex, occupational classification, and house- hold income. For the other models discussed in this chapter which have binary dependent variables, the criterion of a .200 estimated effect or greater was used to choose those sets of charac- teristics which have the greatest effects. The age of the household head exhibits a relationship to the probability that the household enjoys exclusive access to kitchen facilities which first decreases and then increases as age increases from 15 to approximately 65. This first variation in probability stays within narrow limits, a range of less than .006. After age 65 the probability that the household enjoys this desirable housing characteristic decreases at an increasing rate. This relationship is computed and presented in Table V-3. The sex of the household head exhibits an estimated .042 net effect on this desirable housing characteristic. A household with a female head has a .042 higher probability of enjoying exclusive access to kitchen facilities than a household with a male head. The occupational classifications of the household head exhibit a different estimated pattern of relationships than those exhibited with other dependent variables. The classifications are listed here from the one with the highest estimated probability to the lowest: farm service worker, farm foreman, farm manager, farm laborer, farmer, blue collar worker, white collar worker, no 146 occupation, laborers, occupation not reported, and service worker. These classifications represent an estimated .047 probability range from highest to lowest. Household income also exhibits a different pattern of relation- ships with the probability of exclusive access to kitchen facilities than has occurred previously. At zero income households composed of families have the lowest estimated probability, households composed of unrelated individuals next, and mix households exhibited the high- est probability. The relative magnitudes between the logarithm of household income and the probability of this desirable housing charac- teristic, however, is the same as the pattern commonly observed. It is the greatest for mixed households, less for households composed of families, and the least for households composed of unrelated indi- viduals. These sets of socio-economic and locational characteristics have the strongest estimated relationships with this dependent vari- able. These and other relationships may be observed in Table IV—lO. Ylo’ Telephone Available The dependent variable for the last model of this section indi- cates the availability of a telephone. As presented in Table IV-ll, 23.0 percent of variation in this dependent variable is explained by the regressors. Four sets Of sociO-economic and locational charac- teristics exhibit an estimated effect on the probability that the household has a telephone available greater than that of other sets of characteristics. They are: age of the household head, 147 TABLE IV-ll.--Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and Telephone Available . . c Regression R2 Predeterm1ned Variables Coefficient Deletes Constant Term -.l920 Region of the United States X1 Northeasta d .0223 .2296 X2 North Central ---- ---- X3 South8 .1063 .2209 X4 West3 .0318 .2293 Size of Place x Rural Farma .0990 .2296 x6 Rural Nonfarma .1038 .2296 X Urban Territory Outside of Places8 .0075 .2299 X3 Log. of the Size of Place (Population)b .0008 .2299 Location Within Urbanized Area X9 In a Central Citya .0134 .2299 X10 In Remainder of Urbanized Aread ---- ---- Age of Household Head x11 Ageb/lo .2768 .2281 x12 Age Squaredb/1,000 .4859 .2286 x13 Age Cubedb/IOO,000 .2798 .2289 Sex of Household Head x14 Maled ———- ---- X15 Femalea .0520 .2286 Race of Household Head X16 White With Spanish Surnamea .1234 .2287 X Whited --—- ---- 17 a X18 Negro .1201 .2245 x19 Other Racea .0150 .2299 Nativity and Parentage of Household Head X Native With Native Parentsd ---- ---- X Native With One Foreign Parenta .0283 .2296 X22 Native With Foreign Parents8 .0227 .2296 X23 Foreign Borna .0332 .2295 Metropolitan Residence in 1955 X24 Same Housed ---- ---- X25 Different House Same County8 .0183 .2296 x26 Different County Same Statea .0260 .2297 X27 Contiguous State3 .0352 .2298 x28 Noncontiguous Statea .0467 .2294 x29 Abroad or at Seaa .0294 .2299 LO De Sc 148 TABLE IV-ll.--Continued . . c Regression R2 Predetermined Variables Coefficient Deletes OCCUpational Classification X30 Farmera .0199 .2299 X31 Farm Manager: .1042 .2299 X32 Farm Foreman .1433 .2298 X33 Farm Laborer a -.1505 .2287 X34 Farm Service Workera .0060 .2299 X35 White Collar Worker .0829 .2288 X36 Blue Collar Wogkera .0284 .2298 X37 Service Worker .0258 .2298 x38 Laborera -.0519 .2296 X39 Occupation Not Reportedd ---- ---- X40 No Occupationa -.0018 .2299 Tenure x41 Owneda .1481 .2087 X Rentedd ---- ---- 42 X No Cash Renta .0338 .2297 Educational Level of Household Head X44 Educational Level if 510.5 Yearsb .0219 .2199 X45 Educational Level if >10.5 Years .0034 .2285 Log. of Household Income for Households of Various Types X46 Unrelated Individuals (Slope)b .0385 .2290 X4 Family (Intercept)a -.0633 .2298 X48 Family (SlOpe)b .0884 .2221 X49 Mixed (Intercept)a -.3545 .2296 X50 Mixed (SlOpe)b .1650 .2291 Dependency Ratio b X51 Dependency Ratio -.0215 .2287 x52 No One 14 64a -.0008 .2299 R2 = .2299 aThis variable is dichotomous equalling condition holds, zero otherwise. b O 0 O O This variable is continuous. one if the stated cThe observation unit is the household and the variables pertain either to the household or to the head of household. dThis variable was omitted to avoid singularity. Source: One-in-a-thousand sample tapes, 20 percent sample, Censuses of POpulation and Housing [36]. 1960 149 occupational classification, education of the household head, and household income. The age of the household head exhibits a .392 increase in the probability that the household has a telephone available for a lOO-year-old household head over a lS-year-old household head. The relationship, which is presented in Table V-3, first increases at a decreasing rate, reaches a plateau at about age 45 where it decreases slightly, and finally increases from about age 75 on. The occupational classification of the household head exhibits an estimated .294 range between the occupational classification with the highest probability of having a telephone available, farm fore- man, and the one with the lowest, farm laborer. The pattern of relationships with this dependent variable is different from that with any of the other dependent variables and will be discussed in Chapter V. The education of the household head accounts for an estimated .286 increase in the probability that a telephone is available as education goes from zero to five years of college or more. The positive relationship is linear with a greater slope between zero and ten and a half years of education than beyond ten and a half years. The R2 deletes indicate that the percentage of the variation in the dependent variable that is explained would decrease by 1.00 if X44 were removed. Household income exhibits a very familiar pattern of relation- ships with this dependent variable. The relationships between the logarithm of the household income and the probability that a 150 telephone is available is positive with the greatest slope for mixed households, next for households composed of families and the smallest for households composed of unrelated individuals. At zero or negative income households composed of unrelated individuals have the highest probability of having a telephone available, households composed Of families next, and mixed households, the lowest probability. The R2 deletes indicate that two variables, not previously discussed, may explain a substantial proportion of the total variation in the dependent variable. Omitting owned, X from the model would 41 reduce R2 by .021. Owner occupied housing units have a .148 greater probability of having a telephone available than renter-occupied units. The other variable which indicates the South, X and has a negative 3 relationship with the dependent variable, if omitted from the model would reduce R2 by .009. The estimated relationships just discussed and others which appear less substantial are presented in Table IV-ll. Conclusions The ten models discussed in this section have a common set of independent variables. These independent variables included thirteen sets of socio-economic and locational variables which were defined in the first part of the chapter. The ten binary dependent variables, which were also defined earlier, are derived from the measures of housing condition which are included in our INDEX discussed in Chapter II. The purpose of this section has been to examine the estimated net relationships between the socio-economic and locational characteristics of the occupants and measures of housing condition. 151 Specifically we have examined the relationships between those four, five, or six sets of sociO-economic and locational characteristics which exhibit the strongest estimated net relationships with the measures of housing condition. In nine of the ten models examined, household income was among the sets of variables with the strongest estimated net relation- ships. Only in the model with the binary dependent variable, built from 1950 to 1960, was household income not among the sets of primary explanatory variables. The set of variables describing household income was defined to allow for a different intercept and lepe for each of the three types of households--unrelated individuals, families, and mixed. The relationships were generally strongest for mixed households, next for households composed of families, and weakest for households composed of unrelated individuals. Another set of variables which exhibited consistently strong relationships with the measures of housing condition was the occu- pational classification of the household head. This set was listed among the sets of primary explanatory variables for all but two of the dependent variables--six rooms or more and built from 1950 to 1960. The set of variables describing the education of the house- hold head were also frequently among the primary explanatory variables, for seven of the ten models. Several other sets of socio-economic and locational charac- teristics which appeared among the primary explanatory variables less frequently are listed here from the ones which appeared more frequently 152 to those which appeared less frequently: age Of the household head (five of the ten models), size of place (four of the ten models), dependency (three of the ten models), type of tenure (two of the ten models), and race (one Of the ten models). The three sets of characteristics which appear to consistently have the strongest estimated net relationships with measures of hous- ing condition are: household income, occupational classification and education of the household head. The form of these and other relationships with the different measures of housing condition will be presented and discussed in Chapter V. At the same time these estimated net relationships will be compared to the estimated gross relationships presented in Chapter III. In the next section of this chapter, a model, which includes the same explanatory variables as were used in the ten models of this section and the INDEX constructed in Chapter II as dependent vari- ables, is discussed. Net Relationship with INDEX In this section we will be examining the estimated net relationships between the sociO-economic and locational variables defined in the first part of this chapter and the measure of housing condition constructed in Chapter II--INDEX. This model has more desirable properties than the models of the previous section. The error terms for this model are assumed to be both homoskedastic and normally distributed in addition to the desirable assumptions of the previous models. The assumptions of this model are described more 153 fully in Appendix III. The resulting OLS estimates of the regression coefficients are unbiased, efficient, and consistent. The empirical results presented include OLS estimates of the regression coefficients, estimated standard errors for the coefficients, the level of significance at which the null hypothesis that the coef- ficient is equal to zero is rejected, and the R2 delete for each coefficient. As was the case with the ten previous models the R2 deletes are somewhat misleading. Because a certain amount of multi- collinearity exists between the predetermined variables, the R2 deletes are overstated. When an R2 delete is calculated for a particular variable, only part of the effect of that variable is removed. De- pending upon the degree of multicollinearity varying proportions of the effect of the omitted variable are picked up by the included variables with which it is correlated. Several statistics are included which relate to the total model: R2, F, and significance level for the null hypothesis that all estimated coefficients equal zero and the standard error of estimate. Empirical Results The empirical results presented in Table lV—12 indicate that 44.9 percent of the total variation in INDEX was explained by the predetermined variables. This means that more than half of the variation in the dependent variable is not explained. The relation- ship between the predetermined variables and INDEX is significant at <.0005 level of significance. 1541 TABLE IV-12.--Estimated Net Relationships Between Socio-economic and Locational Characteristics of the Occupants and INDEX - 2 . . c Regress10n Standard Level of R Predeterm1ned Variables Coefficient Error Significance Deletes Constant Term 42.7146 1.1226 <.0005 Region of the United States xl Northeasta d 1.0109 .1049 <.0005 .4475 X2 North Central --- --- --- --- X3 Southa - 1.6701 .1025 <.0005 .4452 x4 Westa .5942 .1199 <.0005 .4484 Size of Place a XS Rural Farm - 2.2784 .5122 <.0005 .4485 x6 Rural Nonfarma - .8000 .4835 .094 .4487 X7 Urban Territory Outside of Places3 b 4.6818 .5065 <.0005 .4476 X8 Log. of the Size of Place (POpulation) .8189 .1115 <.OOOS .4480 Location Within Urbanized Area X In a Central City8 d - .5635 .1725 .001 .4486 X10 In Remainder of Urbanized Area --- —-- —-- --— Age of Housghold Head X Age /10 4.9949 .5945 <.OOOS .4478 x12 Age Squargdb/l,000 - 8.7851 1.2153 <.ooos .4481 X13 Age Cubed /100,000 4.8088 .7876 <.0005 .4483 Sex of Household Head X14 Male a --- --- --- --- Xls Female .6401 .1281 <.0005 .4484 Race of Household Head xl6 WhitedWith Spanish surnamea - 2.8069 .3219 <.0005 .4477 X White --- --- --- --- 1 a X18 Negro a - 4.5974 .1491 <.OOOS .4362 X19 Other Race - 3.8722 .6931 <.0005 .4483 Nativity and Parentage of Household Head X20 Native With Native Parents a --- --- -—- --- X2 Native With One Foreign Parent .5540 .1518 <.0005 .4486 x22 Native With Foreign Parentsa .4240 .1215 .001 .4486 x23 Foreign Borna 1.7184 .1496 <.0005 .4470 Metropolitan Residence in 1955 X24 Same Housed a --- --- --- --— X25 Different House Same County 2.0423 .0935 (.0005 .4424 X26 Different County Same State 2.8674 .1571 <.0005 .4443 X27 Contiguous Statea 2.8926 .2409 <.0005 .4468 X28 Noncontiguous State 3.2813 .1812 (.0005 .4444 x29 Abroad or at Sea8 2.8801 .3420 <.0005 .4478 1555 TABLE IV—12.-—Continued . . c Regression Standard Level of R2 Predeterm1ned Variables Coefficient Error Significance Deletes Occupational Classification x30 Farmera a - 1.0870 .3148 .001 .4486 X31 Farm Manager 5.0836 1.6100 .002 .4486 x32 Farm Foreman: 2.9087 1.4351 .040 .4487 X33 Farm Laborer a - 5.3991 .3969 <.OOOS .4463 X34 Farm Service Worker - .2788 2.8897 .886 .4488 xSS White Collar Workeg 1.5713 .2283 <.OOOS .4481 X36 Blue Collar Worker - .0063 .2261 .927 .4488 X37 Service Worker3 - .3585 .2554 .156 .4487 x38 Laborera d - 2.7170 .2737 <.OOOS .4474 X39 Occupation Not Reported --- --- --- --- x4o No Occupationa - .3381 .2643 .198 .4487 Tenure x41 Ownedad 5.8115 .0930 <.OOOS .3971 X42 Rented a --- --- -—- --— X43 No Cash Rent - .9059 .2116 <.OOOS .4485 Educational Level of Household Head b X44 Educational Level if $10.5 Years .7527 .0200 <.OOOS .4301 X45 Educational Level if >10.S Years .1018 .0081 <.OOOS .4467 Log. of Household Income for Households of Various Types b X46 Unrelated Individuals (Slope) .7579 .1129 <.OOOS .4482 x47 Family (Intercept)a - 2.1757 .4796 <.OOOS .4485 x48 Family (Slope)b 2.5145 .0914 <.OOOS .4387 x49 Mixed (Intercept)a -14.3745 1.9147 <.OOOS .4480 x50 Mixed (Slope) 5.7945 .5108 <.OOOS .4470 Dependency Ratio b xSl Dependency Ratio - .1677 .0564 .003 .4486 x52 No One 14-648 .4948 .1764 .005 .4486 R2 - .4488 F . 770.0456 Significance Level - <.OOOS Standard Error of Estimate - 7.6206 8This variable is dichotomous equalling one if the stated condition holds, zero otherwise. bThis variable is continuous. cThe observation unit is the household and the variables pertain either to the household or to the head of household. dThis variable was omitted to avoid singularity. Source: One-in-a-thousand sample tapes, 20 percent sample, 1960 Censuses of Population and HouSing [36]. 156 Most Of the individual coefficient estimates are signifi- cantly different from zero at <.OOOS level of significance. Only four estimates are not significantly different from zero at <.05 level of significance: farm service worker (X34), blue collar worker (X36), service worker (X37), and no occupation (X40). All of these variables are within the set called occupational classification. Thus even though less than half of the variation in the dependent variable is explained, most variables have an estimated effect which is significantly different from zero. The sets of variables which are the primary explanatory vari- ables as would be expected are the same as those for the ten models of the last section: household income, occupational classification, and education of the household head. Four other sets of character- istics have relatively strong relationships to INDEX. They are listed here from the sets with the larger to the sets with smaller relationships: size of place, type of tenure, race and age of the household head. The direction and magnitude of these relationships will be discussed in Chapter V where comparisons will be made be- tween the relationships exhibited in this model and the ten models of the previous section. In the next section Of this chapter, the INDEX will be examined for weight sensitivity. This process will provide some information about the validity of the estimated coefficients for the model presented in Table IV-12 as well as the validity of INDEX as a measure of housing condition. 157 Weight Sensitivity of the INDEX Ideally we would like to examine the basic question of how well INDEX measures housing condition. However, we will instead examine this INDEX for weight sensitivity. More specifically we will be asking if the INDEX is weight sensitive, as used in this study. The INDEX is the regressand in a multiple regression model (previous section of this chapter) which is used to estimate the net relation- ships between thirteen sets of sociO—economic and locational charac- teristics of the occupants and housing condition. The question is, "Are the estimated net relationships dependent upon the weighting system used to construct the INDEX?" Procedure In order to examine this question, we use the regression model of the previous section. Twenty different weighting systems are used to construct twenty versions of the INDEX. These are then used one at a time as regressands with the common set of socio- economic and locational characteristics of the occupants as regressors. The differences in parameter estimates are used as an indication of the sensitivity of the INDEX to weight changes. Hypothesis We assume that if the INDEX is not weight sensitive that the parameter estimates will change very little as weighting systems are changed. This hypothesis creates a measurement problem. How much should a parameter vary and how many parameters can vary before the INDEX is judged weight sensitive? 158 A measure was developed to describe the relative size of the range in parameter estimates. For a Specific range in parameter estimates M - M x : R/M (Ev + M )/2 where: Mx = the maximum parameter estimate, and Mn = the minimum parameter estimate. R/M is the absolute value of the range divided by the mid-point of the range. Thus it measures the size of the range relative to its mid-point. This is justified on the assumption that the approximate importance of a certain size range is inversely proportional to the absolute value of the parameter estimates. R/M does not provide any answers as to how large a change should be tolerated but does provide a way of measuring the relative change. The Models We will next discuss the models used to examine the question of weight sensitivity. First, the dependent variables used in the multiple regression models are presented, then the predetermined variables. Endogenous Variables Twenty different sets of weighting systems are used to form twenty different regressands, INDEX 1 through INDEX 20. The measures 159 of housing condition and the values allotted for the levels of hous- ing condition are presented in Table II-3. INDEX 1 through INDEX 20 are constructed as follows: INDEX 1 = Z W.V. where: i = the ith weighting system for the INDEX. j = the number of the condition measure as listed in Table II-3. V. = the value allowed for the jth condition measure as J listed in Table II-3. W. = the weight given to the jth condition measure. The weighting system for INDEX 1 through INDEX 20 are presented below. INDEX 1 W1 = 0 wj = 1 j s 1 INDEX 2 N2 = 0 WJ. = 1 j g 2 INDEX 10 W10 = 0 Wj = 1 J 2 10 INDEX 11 W1 = 3 Wj = 1 j x 1 INDEX 12 w2 = 3 wj = 1 j 4 2 INDEX 20 W = 3 W. = 1 j = 10 10 j 160 In INDEX 1 through INDEX 10, referred to as the first group, each of the measures of housing condition is set to zero in one INDEX. Each INDEX i represents the sum of all measures of housing condition, except the ith. In INDEX 11 through INDEX 20, referred to as the second group, the value assigned for each measure is multiplied, one at a time, by a factor of three. For this group, INDEX i is three times the (i - 10)th measure of housing condition plus the sum of the other measures. These twenty different weighting systems are used to form the twenty regressands, INDEX 1 through INDEX 20, used in the models here. Predetermined Variables A common set of predetermined variables is used for all twenty models. This set is with two exceptions identical to the re- gressors presented in the Model Specification section of this chapter. The two differences have to do with the specification of the income variables and the education variables. Education is described by two binary variables: X44 = the number of years of formal education if less than or equal to 10.5 years 0 otherwise x45 the number of years of formal education if greater than 10.5 years 0 otherwise This error in the specification of X44 results in the esti- mation of two linear relationships between housing condition and the education of the household head, one for less than or equal to 10.5 years of education and one for greater than 10.5 years of education. This is not a realistic specification because both relationships are 161 forced through the same intercept. The specification discussed in the section, Model Specification, provides for a continuous relation- ship with a change in the slope at 10.5 years of education. This error in specification means that the parameter estimates are not valid descriptions of the relationships between education and housing condition. However, it is believed that they will provide information on the weight sensitivity of the INDEX with respect to education. The other specification error in these models involves the variables used to describe income. X47 and X49 have been omitted from the models. As was discussed in the section, Model Specifi— cation, the intention was to allow for a slope and intercept differ- ence in the log-linear relationships between household income and housing condition for each type of household. The omission of these two binary variables forces the relationships for each of the three types of households through the same intercept. This error in specification means that the resulting parameter estimates are not representative of the structural re— lationship between household income and housing condition. However, as is the case with the education variables, the parameter estimates are believed to provide information on the weight sensitivity of the INDEX with respect to household income. The assumptions regarding the error terms for these models and the properties of the ordinary least squares estimates of the regression parameters are presented in Appendix III. The parameter estimates are assumed to have the desirable small sample and large sample properties. 162 Empirical Results The parameter estimates presented here are the maximum and the minimum estimates for each variable for INDEX 1 through INDEX 10, the first group, and for INDEX 11 through INDEX 20, the second group. Also, R/M has been calculated and is presented for each maximum and minimum. Each model produced at least one maximum or minimum. Two models produced three maximums or minimums and the other sixteen models produced six or more maximums or minimums. The empirical results are presented in Table IV—l3 and Table IV-14. Six of the thirteen sets of socio-economic and locational characteristics contain variables whose parameter estimates are sensitive to weight changes in INDEX. (1) Among the region of the United States variables, West exhibits a parameter sign change with the first and second groups of indexes. South exhibits a parameter sign change with the second group of indexes. (2) Among the size of place variables, the rural nonfarm residence category exhibits a parameter sign change with the second group of indexes and a R/M value of 1.998 with the first group of indexes. (3) Among the location within an urbanized area variables, in a central city ex- hibits a parameter sign change with the first group of indexes and a R/M value exceeding one with the second group of indexes. (4) Among the occupational classification variables, two exhibit parameter sign changes with both the first and second groups of indexes, farm service worker and blue collar worker. The variable, no occupation, exhibits a parameter Sign change with the second group of indexes. The variables, farmer and service worker, both exhibit R/M values New. mmeo.s - o~eo.e woe. amma.n mmwe.~ - aooaz uoeoo ”fix ONA. oeoo.o - mnom.e «NA. eono.e Nama.n - aoauoz sax o o o o o o cause: on mom. ameo.e - aamm.~ Nae. nmoo.~ eefla.s - eoeacaam gmficeom cum: cams: x use: emonomso: mo comm NNH.H amen. onn.H moo. snow. aooo.~ aosaaoa ”fix o o o o o o case: x vac: vaonomsox mo xom ooo.~ oooo. Loco. ooo.~ oooo. Hooo. oeooso oua wa ewe. nmso. - mace. anm. ~eoo. nmoo. - oeoaaaam ou< HHx oHo. nose. name. woe. «on». eefim. oou< x new: vflonomaom mo ou< o o o o o o eaoa< poemcaoaa mo aoocaean cu owx meo.H mesa.“ - mmms. osmo.n oeao. amefi. «some Aesocou a a“ x mou< vaufieenua away“: :ofiuuuoq «so. omae. meme.s fine. some. Hfiom. oncosoasaooav ooesa mo oumm on» me .uos Mx com. «sad.» Ammo.a mam. nono.~ emnn.m aoeaaeao xaooaaaoe some: ox ”m enme.n eo~o.~ - «was. was.” “035.3 Rooo. - eaaamcoz Hausa mx .1 m~n.~ mosm.e - case. use. come.» eons.s - aaaoa seas: x comma mo ONMm oo~am.m Nome.“ - aama.~ once.” cane. once.“ «one: mg omom.~ nono.m - Hmom. emu." eooo.~ Nae". - acosom Nx o o o o o o causecou geaoz ax ”do. ANmm. woae.~ Nae. Rhea. aoo~.H eunoogeaoz x moumum vouwCD 0:8 MO Gawuox ..ooe. Neeo.ae mNem.Ho mam. exam.~n Hooa.~e ence ocooacou once: we omcmm mo Ochoa-eaz enema“: aassxaz echoa-omz ensue“: aaaaxaz \euedm \oucam ON xmazunnua xmozm nacho veooom oH xmoznuna xmazH macho umuwm UmOHdeHm> fiOGwEHOHOGOHA amueamsuuo.o:u mo muwumwuouoauunu Hmcofiumuoq one uflaoeou016woowv m u novcH :owufivcoo .munuwoz nova” «caucasu new: mauve: sneeze on» scum moueEMumm nouoamuem cowmmoumom camwufisz assflcflz use asafixazii.n~->n m4mos secoaeaoaem ”fix can. aeNA. noes. no”. osmm. mama. o.aaa m.osv we Ho>os "acoseeoaem x VGO: UHOSOMfiOI MO ~0>0J HGCOfiUdUSVN boos.~ oooo.~ Noos. can." coon.“ ewes. aocoa coco oz mux o o o o o o eeoocoa Hex ofle. oooa.m oNOH.m nan. AH-.4 Nooa.m aoocso x eunuch o-e.v some. aesn. can." eewe. noes. acoaeaoaooo oz me o o o o o o ovouuomom uoz coflummsuuo x mom. o~ao.n oooa.~ - no”. moma.~ aoa~.~ aaoaooes mmx nHH.~ naoo. ewes. - mom. «nae. Amos. aaoxaoz ooa>aom onx oo~m.o Hans. eonN. cams. N mooo. eons. aaoaaoz aessoo ossm max doe. oma~.~ once.” new. name. ae~n.~ «cameo: aessoo been: nx o~o~.aH oa~o.~ anoo.~ ooss. ms meme. nm~o.s aaoaaoz oomsaom eaaa nnx nan. o~na.a enam.m - com. Hmem.m ooa~.e aaaaooas seam ~nx ~No. ome~.~ on~.e van. emeo.~ eamo.n eceeoaoa seam Hex an». eaNH.m Nomo.a axe. aoam.n maem.m aaouacez sued omx moa.s dame.“ noNH. - nee. Hmeo.s none. aaoeaaa x cofipeowmfimnmgu ”macaumgsouo 4mm. oeoo.~ mo~m.m NNo. memo.“ cama.~ aaom ea no eaouo< aNx an». monm.~ neom.o Ada. mae~.s woe~.n aooaom nsoauaocoocoz mwx AN». Haw~.~ nnsm.m nae. ooa~.s «sam.~ eooeom msoausocou MNx one. NNam.N mama.m ova. eHaH.s Loom.m aooaom oaam socaou economon mwx mam. Aeoo.~ mmee.e Nmo. amen. Neos.~ axocaou seam ease: economon Nx o o o o o o noose: oaam x mmaa ca oucovumom ceuamonouuo oz can. meow." oa-.~ ~m~. omen.s o-a.s eeaom cumoaoa mmx oes.a «mew. seam. Nee." oees. ~Nme. anecoaea cueoaoa cos: o>aoaz HNX mmn. anew. coma. om~. onae. nsoo. «oedema cusoaoa oco com: o>soaz oNx o o o o o 0 bureaus o>aoaz cos: osaeaz x poo: vaocomsoz mo omaucoumm van >uw>wumz cued: we once: we ocaoa-eaz assess: eaamxe: oesoa-oaz eases“: aaemxaz \oucam \oucam umoHneflum> vocfisAOuovoum om xmoznuaua xmazm macaw vacuum cu xmnzmitfi xmnzH nacho umuflm voscfiu:0011.n~u>m mqmm on wouuwao mm: oHnmfium> mankv .uaosomso: we can: one on no vaonomao: one on nonpflo cfimuhom moflnmfiua> on» can eHonomso: on» mfl was: cofipm>uomno echo .mzoacwueou mw mapmfium> macho .emwauonuo ones .mvuo: :owuwvcoo woueum ecu m« use unwfiaesve mnoaouonuwu m“ manefium> mwghe ano. eaoe. oes~.s one. ooom. oeoo. aeo-eo oeo oz mmx onso.o mace. - memo. omo.s oon. - Hooo. - ooeooz zocoocoooo x ofluam zucovcomon ooN. moom.~ meow." ooN. aomo.s Hoe~.~ ooxmz mmx oem. oooe.~ mooo.n Nos. mono.z HHA~.~ o Hosea cox Non. memo.H omom.~ Hem. ammo. mooz.s oosooom>moco oooezoaeo x monxh anewue> mo mvdonomsoz new oaoucH vaonomaoz mo .moa omcmx once: mo anaemic“: adage“: asawxuz ucfiomuvwz snags“: ESEMxmz \oucmm \owsmx omoHnmfinm> vocwEhouououm on xmnzuiugfi xmozH OH xmnzHuiu xmnzH macaw vacuum guano woman vozcfiucou-i.mH1>H wqmHooHoeaoo< ooH o.oH N.N o.N o.N o.oN o.HH o.mN o.HH momcoz No ocooaoa ow NH N o o NH oH NN oH nausea mo .oz Hooch o.mH o.oN o.mN o.om N.om N.oo N.No o.ooH amoeoooeoa o>HooHseaoo< ooH o.oH m.N N.o m.o m.mN o.HH o.NN m.N oomcom mo ecooaoa me o H N o oH m NH H ooNcoz mo .oz oN-HH zmozH macaw vacuum m.o o.HH o.oN o.mN N.eo N.mm H.oN o.ooH omeecooaoa o>HooHoeaoo< ooH m.o N.N m.o N.o o.oH o.HH m.mN o.oN newcoz mo accused me e H o N o m oH o nomcez mo .oz oH-H HmozH macho umhflm N NW N ow cmc H cmca . . . . Hoooe o N .m o N .m W H m H H H H ma H m. H mN H z\z oo ooze oo coco m H mo H mo ma mo m mo mN mo o mo HopoEmHmm Houosmnmm oz osHm> z\m NHOzH x9 mH1>H mHan Eoum momcmm HouoEmumm one mo coflpznfluumfio xoconvonmii.eHu>H mamH moHnmh Eopm coxmp mu:OHOHmmooo cowmmoumou on» one maficmcoflumHou um: nonmeflumo omega oH 4H xmozH oco H->H oHoae cH ooeconoea one N - soum meonmoumou on» How moHpmHHm> unaccomon xumcfin on» one cofiufivcoo mcflmson mo mousmmos omonhm Hz moHoaHao> .NH1>H "boasom .HH aoooeco oH oochoo .coxmu one mammauwmmooo woumEHumo on» gowns Neom. HoNo.H- ooHo.H xmozH Hmo.- mooH. - mNNo. - oHooHHa>< ococooHoe oH» HHoo. omoo. Nooo. ooHOHHHon coHoHHz oo amooo< o>HooHoxm oz moNH.- NNom. - oomo. ocoeoHaom NcHoooz oz oNoo. mNoo. - ommo. meooacoam oaoz so ago N» Homo. mmoo. oNHo. - oooH 6» omoH eoaa HHHoo oz mono. mNmo. - ommo. Noaozm no zoom a on noooo< ooHooHoxm m» emNo. oooo. - ooNo. ooncH oooHa noooz oHoo oco ooz 4» mooo.- NHHo. - NNoo. oooooHooHHo xHHoaaooaaem ooz m» oHHo. oHNo. - ooHo. ocaom NHHeaooooaom Nz oooo.- omeo. - ammo. oaoz so neoom me Hz Hooz ox coaom Hz mewm Hx moumum umuflca ecu mo mconom mcoHuHucou mchso: mo monommmz coHoHocoo mchso: mo moASmmoz cam moumpm nOuHc: one mo mcofiwom 2663pom mmwnmcoHumHom uoz ecumeflumm--.Hu> mHmH nmsonnu N1>H moHan Eoum coxmp mucOHOHmmooo coflmmOHMON may one mchmcoHumHmH uoc poemswumo omonh Hoousom .HH Hoooeco cH ooeHmoo ma xmozH cam H1>H oHan :H poucomOHm one OH» 1 H» moHpmem> .coxmu mum mucoHOHmmooo nopmefiumo on» noflnz Eonm mcofimmopmop onu pom moHanHm> unopcomow xpmcfip ago one :oHancoo mcflmzo: mo mouammoe omonhm 178 HHoo.- oNHN. HHoo.o oooo. - eNNN.N- zmozH omHo. oooo. HNoo. NmoH. - oooo. - oHooHHo>< ococooHoH on mHoo. mmoo. - mmHo. - NHHo. - HeHo. - ooHoHHHoaa cocoon m CH WWOUU< 0>HW3HUXW V Nooo.- mmoo. HNoo. ommN. NNoH. ecoeoHoom NcHoooz oz oHoo.- Nomo. NHHH. NNHo. - eomo. - noooacoao one: so oco N» emNo.- omHo. NHHN. oNNo. HNNo. - oooH oo omoH some HHHoH oz oNoo. ooNo. Nooo. Nemo. - oooo. - aosoem no comm m w op mmooo< o>HmsHoxm > omHo. NNNo. oNHH. mmNo. NNmo. - ooncH ooaHa e aoooz oHoo oce oox z HNoo. omHo. oomo. NoNo. Hooo. oooooHaeHHo NHHoaaooouom Hoz m» oNNo.- memo. ooNH. oooo. mmoo. ocaom xHHoaooooaem NH NmNo.- omNo. - mooH. - oNoo. - Nemo. one: so msooz me H» xuwu oomHm mo ONHm moomHm oufimuso Hanucou ecu mo mofipouwpuoh Summcoz Show w :H ax Enufihmwoq mx copy: ox Hausa ox Hausa mx cowuwpcou mcfimzo: mo monommoz m oomHm mo ONHm cofiufivcou mcflmso: mo mousmmoz new mom< poNHcmnH: cm canHz :oHmeOH ecu oomHm mo oNHm coozuom mmfinmcowumaom uoz woumafipmmuu.mi> mqmH swoops» ~->H moHnmh scum coxmu mucoonmmoou :onmouuou on» one mnwnmcowuuHou no: voumaHuno cough "ouunom .HH Nounenu :H vocfimev «H xmozu new H1>H oHnuH eH voucomOAQ can an» 1 H» noapauua> .coxau one mucoHowuuooo voumEHumo ecu seas: Scum meowmmouuou any new moHneHum> ucovcomov xumcfln ecu eke acouwvcou newnson mo nounnaoa omoghd coH.oH NoN. oco.- ooN. oNo. oNo. oNo. HNH. cNo.- oNo.- Hoo.H ooH ooo.o ooc. Noo.- ocN. ooo. ooo. ooo. coH. HNo.- NNo.- Noo. co Hoo.o moo. ooo.- omN. Noo. oNN. ooo. ooo. oco.- oNo.- ooo. oo cHo.c oco. coo.- ooN. ooo. ooN. coo. oco. oco.- cNo.- Noo. no coo.o Noo. Hoo.- HoN. coo. oHN. coo. cNo. HNo.- oNo.- coo. oo ooo.c oNo. NNo.- ooN. ooo. ooN. co-. oNo. oNo.- Noo.- oNc. oN HHo.o cHo. oNo.- NcN. coo. oHN. ooo. oco. NNo.- Noo.- oHc. oN coo.o cHo. oNo.- oNN. ooo. ooN. Noo. Hco. ooo.- Noo.- cHo. oc ooN.c cHo. oNo.- oNN. Hoo. ocN. ooo. oco. oco.- Noo.- oHo. oc ooc.o oHo. oNo.- coN. NNo. coN. cNo. coo. Noo.- Noo.- oHo. co oNo.o oHo. oNo.- HoN. oNo. oNo. NNo. ooo. ooo.- ooo.- ooo. oo oco.o NHo. oNo.- ooN. oHo. Noo. oHo. Noo. Hoo.- ooo.- NoN. oo Hoo.o ooo. oNo.- NoN. oHo. Hoo. oHo. ooo. Hoo.- Hco.- oNN. oo NcN.c ooo. cNo.- ooN. HHo. Noo. HHo. ooo. ooo.- Noo.- coN. co cNo.o oco. cNo.- NNN. ooo. Hoo. Noo. ooo. ooo.- Hco.- HoN. oo ooN.N Noo. oNo.- NoN. ooo. ooo. ooo. Hoo. ooo.- ooo.- ooc. oN occ.c Noo. NNo.- oNN. Noo. oco. Noo. coo. HNo.- Noo.- oco. oN oNc.c oHo. oHo.- ooH. ooo. oHo. ooo. ooo. ooo.- ooo.- coo. oH choHHo>< oonwwme“ ocosoHaom meooaceoo ocoH aanmum no “WMMMH cocooHooHHo oesoc on“: onunmaoz xmozm ocozmoHoh cu mmuoo< ucwumo: uwumuo souMuummM“ ou mmoou< noun: AHHaumwmauum xHHeuauunuum mace: on» we o>fim=Hoxm . u>mm=Hoxm vHou a no: xflm ou< oH» oz o» N» c» o» o» o» Nz H» NMHx mu. x . aeoHqucou newnso: mo cayenne: coHqucou ucwmaoz mo nounmaoz ecu vac: nHonomao: an» we ou< coozuon neoHueHom no: no.. 183 One pattern of relationships is exhibited with four of the desirable housing characteristics and with INDEX. As the age of the household head increases, this pattern of relationships first in- creases at decreasing rate, levels off or decreases and then in- creases. With the desirable housing characteristic, six rooms or more, the pattern of relationships does not decrease between the ages of 50 and 75 but increases only slightly. For the dependent vari- able, telephone available, the relatively flat portion of the relationships is between the ages of 45 and 75. With the dependent variable, heating equipment, the pattern of relationships decreases between the ages of 45 and 85 increasing only slightly beyond that age. With the dependent variables, built from 1950 to 1960 and INDEX, the pattern of relationships decreases sharply in the middle range of ages. This estimated pattern of relationships indicates that the age of the household head is positively associated with these measures of housing condition for the young household heads and the old but has a negative or no effect on housing condition for the middle range of ages. Another estimated pattern of relationships was exhibited with the dependent variable, hot and cold water piped inside. The pattern has a curvilinear form which is almost linear. As the age of the household head increases, the probability that the housing unit has hot and cold water piped inside increases. The third pattern of relationships occurs with the housing characteristics, structurally sound and not structurally dilapidated. As the age of the household head increases, the probability that the 184 housing unit has the desirable housing characteristics first de- creases until about age 35 and then increases until about age 90. Two other patterns of relationships are observed. The probability, that a housing unit has either exclusive access to a bath or shower or one or more bathrooms, increases as the age of the household head increases to about 80 years of age decreasing thereafter. The last pattern of relationships is observed with the dependent variable, exclusive access to kitchen facilities. The probability that the housing unit possesses this desirable housing characteristic first decreases slightly to about age 35 then in- creases slightly to about age 60 and decreases at an increasing rate thereafter. The variety in patterns of relationships between various measures of housing condition and the age of the household head is surprising. The "range of effects” for estimated net relationships with three of the housing characteristics exceeds .300--six rooms or more, not structurally dilapidated, and telephone available. With four of the housing characteristics, the "range of effects" is less than or equal to .044--Structurally sound, exclusive access to a bath or shower, one or more bathrooms, and exclusive access to kitchen facilities. With the other three housing characteristics-- built from 1950 to 1960, heating equipment, and hot and cold water piped inside-—the "ranges of effects" are .220, .110, and .091, respectively. 185 Comparison Between Estimated Gross and Net Relationships When comparing these patterns of estimated net relationships (Table V-3) with the estimated gross relationships (Table 111-6), the results are mixed. The desirable housing characteristic, structurally sound, exhibits an estimated net pattern of relationships which is almost opposite the estimated gross pattern. The housing character- istics, six rooms or more, built from 1950 to 1960, and heating equipment, exhibit net patterns of relationships which are similar to the gross patterns. With the other housing characteristics, the gross and net patterns of relationships vary substantially. In general, the "ranges of effects" for estimated gross relationships are greater than the ranges of effects for estimated net relation- ships. However, the reverse holds for the housing characteristic, six rooms or more. In general, the hypothesis that the old and young will experience difficulty in the housing market and will have lower levels of housing condition is not supported by the estimated net relationships. These estimated relationships do indicate that the young are more likely to have lower housing conditions and that the old are more likely to experience higher levels of housing condition than the household heads in the middle age categories with the effects of other characteristics removed. Sex of the Household Head The estimated net relationships for the sex of the household head are presented in Table V-4. The binary variable indicating a 186 .NH->H omoouco mu>H moHan Eoum coxmu mu:OHOHmmooo COHmmoamon esp one mmfinmcofiumHou uoc poumefipmo omogh Hoohsom .HH sooooco cH cooHcoo cH xmozH coo H->H oHcoH cH ooococoaa one oH» - H» coHcaHaoo .oozoo oae cocoHoHccooo ooooeHooo one onos Eoum mcoflmmOHMON on» Now moHanHm> pcopcmmon xpmcfin on» one coHqucoo mcfimso: mo monommoe omogem Hooc. HmozH oNoo. oHcoHHe>< ococooHoe on nHoo. moHuHHHomm cocoqu on mmooo< o>HmoHoxm m» mnoo.- ucoemfiscm wcfiumo: w» mnoo. mEooucumm who: No one u» mNoo.- ocoH oo oooH eoaa oHHoo c» mmmo. Nozosm no comm m ou mmooo< o>wmsHoxm m» ocHo. ocHocH cooHo Nope: oHoo coo co: .oz Nooo. cooacHooHHo xHHoaoooouoc Hoz o» NNoo. coaoc zHHoaaoooaom N» MMNO... 080: .HO mEOO“ Kunm HW 2%” so coflufivcou wcwmso: mo monommoz coo: uHogomsom mo xow :oflufivcou mcflmso: mo monommoz can poo: nHocomso: esp mo xom cooauom mmfinmcoHuwHom poz woueafiummiu.eu> mamH nmsongn m->H moHan Eonm coxmu mucon6nmmooo cenmmonmon on» ono mmnzmconumnon uoc woumEnnmo omonn Hoonsow .HH nonooco on coonooc on xmozH cam H1>H onnoh an woncomonm ono 0H» u H> monnmnno> .coxon onm mucononmmooo poumEnumo onu gong: Eonm mconmmonmon on» now moHnonnm> ncopcomoc xnmcnn on» onm ooHunvcoo wcnmso: mo monzmmoe omonkm 189 NNNo.o- oeoo.o- ocoo.N- xmozH ooHo. - HoNH. - ooNH. - oHooHHono ococooHon oHn oooo. - ocoo. - oooo. - ooHnHHHooa cocoonx on ooooo< osnooHoxm on ooHN. - Hcoo. - NnoH. - neoeoHoom mcnnoox on oHoo. - oocH. - Noco. - oaooncooo one: no oeo Nn Hcoo. omoo. - HHoo. - ocoH on oooH sona nHHao cn oooo. - oooH. - oHoo. - nosocc no zoom o oo oooooo o>HoaHoxm on oooo. - ooNH. - oomo. - ocHocH ooono nonoz oHoo ooo no: on oooo. - oHoH. - HoNo. - cocoonooHHo zHHononoonoc noz on ooNH. - noNH. - cNHH. - oeoom nHHononoonnc N» NHoo. ocHo. - oooo. - one: no oeooo xno H» ooo oeocnsm a oH onwoz on zonoooo can; nocno x ooH cH .n: x con H m m .n.p:ou :nmso: mo monzmmoz poo: cnocomso: mo oomm conpnvcou manmsom mo monommoz poo poo: pnocomso: on» mo oomm coozuom mangm:0nuonom uoz boomsnummui.mu> m4mH :msonnu N->H moanh Eonm coxmn mucononmmooo canmmonmon oz» one an£m¢0nanon no: couoEnumo omonk “oonsom .HH nonoooo on coonnoc on xmozH cam H1>H onnoh an noncomonm ono OH» 1 H» monnmnno> .coxmn ono mucononmmooo woumEnumo ozu sons: Eonm m:0nmmonwon ocu now monncnno> ucoucomov xnmcnn on» onm cenunvcoo manmso: mo monomooe omonho ocHN.H ooNo. ocoo. xmozH Nooo. NNNo. ooNo. oHHoHHo>< ooooooHon oHn oooo. - oNoo.- oooo.- ooHnHHnooo oooonno on oooooo o>nooHonm on oomo. Nooo. ooNo. noooonoom monocoz c» como. HHoo.- NNHo. oooononoo one: no ooo Nn cooo. oooo. ooHo. ocoH on oooH oona nHHoo c» oomo. oooo. ooHo. noaooo no onoo o on ooooo< o>noaHoxm on cnoo. NHHo.- ocoo. ocHooH coona none: cHoo coo nom on ocHo. oooo. oooo. conocnooHHo nHchanoonnc noz on occo. HoNo. ooNo. cooom nHHononoannm N» cooo. - cHHo.- cNoo.- onoz no ooooo xnc Hn muconmm muconmm nooncm :mnonom omHonoo onnz omnonoa ooo ooH: :mnonow mmx nun: o>nuoz mmx o>nnmz me ocoHunccou mcnmSom mo monamooz poo: cnosomsom mo omouconom cam xun>nnoz acnunvcou mcnmsom mo monommoz cam coo: cHogomsom on» mo omcnconcm vac zpn>nncz coozuom mannmconumHom noz monoEHnmmui.ou> mqmH moanen Eonw coxen mucononmmooo :onmmonwon onu one muncmconnenon poc ooueennmo omogh mn xmozn one H->H onnen an concomonm one on» - a n monoonno> .NH:>H ”oonsom .Hn nonoooo on ooonnoo .coxeu one mucononmmooo ooueEnumo onu sons: Eonm m¢0nmmonmon onu now monnenne> ocoocomoo nnecnn on» one connnocoo mcnmoo: mo monomeoe omoghe Homm.m mnom.m omow.~ onew.~ mmvo.~ xmozn ooNo.- ocoo.- Nmmo.- ooNo.- mono.- onpenne>< ococaonoh on» «ooo.- «ooo.- oHHo.- mmoo.- omoo.- monunnnoem cogounx m on mmooo< o>nmsnoxm w Nnmo. oomo. vwno. Homo. ooeo. pcoeansom menueom m» ommo. vmmo. oweo. movo. nomo. meoonnuem onoz no oco 5» nmom. oonm. mmmm. NooN. omnm. ooon on omon Eonw unnnm o> ammo. moeo. Homo. ommo. mmmo. nozonm no suem m e on mmooo< o>nmsnoxm > Hmmo. mnoo. ovoo. nvvo. memo. oonmcn ooano v nopez onou one no: > nono. oomo. nnno. nono. onno. oonoonoonno m xnnenSuoznnm noz > unoo. osmo. ammo. sumo. wmmo. oesom xnnenSHosnnm N» wooo.- evoo.- nmvo.- Homo.- vmvo.- onoz no meoom Knm H> eom ouepm oneum oEem nucsou we no moosMnucou opeum nucsou oeem omSom oeonn< omx -coz wmx mooSMnucou nmx neonoMMno omx neonommHo mmx econunocou mmon an oocoonmom ceunnomonnoz mcnmso: mo monumeoz I] conunocou wcnmsoz mo monomeoz one mmmn on oocoonmom ceunnomonuoz ocn coozuom mannmconuenom uoz ooueEnumm--.n-> m4m< oooooonon onn mooo.- mono. oono. mono. mooo. monnnnnooo coconnx o OH mmmuu< 0>flm3HUXm > mnmo.- omno. - mono.- mono. wooo. nooeonooo eonnoo: on nono.- ommo. nooo. omon. mono. oeoongnom onoz no ooo n» ooNo.- onoo. - mono.- mnon. - mono. ooon on omon gone nnnom on mnno.- onno. mono. onnm. nmno. nozonm no cuem e on mmooo< o>nmsnuxm m» mnoo. Nooo. oooo.- noon. mooo. oonoon ooono o none: onou one no: > nono. ommo. mooo. Mnno. oono. ooooonoonno nnnonoooonnm noz m» «ooo.- oomo. - mooo.- wooo. nmmo. ocoom nnnononoonnm Nn mooo.- omoo. «noo. - nmmn. onno. onoz no oooom nnm n» noxnoz ceEonom noxnoz nomecez noxnoz oon>nom nmx anon mmx nonnoo om anon nmx nonnoo mm . oonm x onnoz x ooonnnoooo mcnmzoz mo monomeo: oeom ononomso: onn now conueonMnmmeHu Heconuemaooo conunocou mcnmnox mo monomeoz one oconueoanmmenu HeconueQSUoo cooznom mangmconnenom poz woueEnumm--.mu> mgmH nononnn ~u>H moanen Eonm noxen mnnononmmooo nonmmonmon onn one mannmnOnnenon non noneennmo omonh ”oonnom .HH nonnono on oonnnoo on xmozH one nu>H onneh nn nonnomonm one on> - H> monnenne> .noxen one mnnononmmooo noneEnnmo onn gonna Eonm mnOnmmonmon onn now monnenne> nnoonoaoo xnennn onn one nOnnnonoo wnnmno: mo monomeoe omonhe 199 noom.m- won~.- oneo.n- onnn.~- nwmm.- noozn moon. - oooo. mono. onmo. - enoo.- onoonno>< ooonoonon on» omno. mnmo. nono. nmoo. nnoo. monnnnnooo oonoonn o on mmouo< o>nmnnoxm > nnnn. - omwo. memo. - Neon. - mono.- nooeonsom monnoo: on noon. - oomo.- nomo. - omno. - om~o.- osoonnnom onoz no ono n» Homo. - onoo.- Nnoo. - onoo. - mnno.- ooon on omon none unnoo on omen. - nnmo.- nnoo. - nnoo. - onno.- nozonm no m :nem e on mmooo< o>nmsnuxm > omon. - onon.- momo. - onno. - ocoo.- oonoon ooono o none: onou one no: > nneo. - omNN.- oeoo. - nnmo. - oooo.- ooooonoonno nnnononosnnm noz on nnmn. - oooo.- mNoo. - mnoo. - enmo.- oooom nnnononosnnm on nono. - omoo.- noon. mnoo. - memo. onoz no onoom nnm n» noxnoz nonmmmw mmx oonwmmw vmx noenem omx mnononen wmx nonnemnowm ovx enonnnonou mnnmno: mo monumeoz oeom ononomno: onn now nonneonMnmmeHu nenonnemnooo voscnncouuu.mu> mam<fi 200 An examination of Table V-8 reveals mixed patterns of esti- mated net relationships. In fact, the pattern of relationships be- tween housing condition and occupational classifications is different for each of the measures of housing condition used. With the de- pendent variable, INDEX, a pattern of relationships is exhibited which appears to be representative of the other patterns. The occupational classifications are listed here from the one with the highest level of housing condition to the one with the lowest: farm manager (X31), farm foreman (X32), white collar worker (X35), occupation not reported (X blue collar worker (X36)’ service 39) ! worker (X37), no occupation (X4 farmer (X farm service worker 0) . 30). (X34), laborer (X38), and farm laborer (X33). Examining the classifications which have the largest positive relationships with housing condition, we see that the effect of the household heads being a white collar worker is always positive over having the classification, occupation not reported. Having the farm manager classification has a positive effect on housing condition for all measures except the year built. In this case, household heads who are farm managers have the lowest probability of residing in a housing unit that was built from 1950 to 1960. At the other end of the spectrum, we see that laborers and farm laborers negatively related to housing condition for all measures except exclusive access to kitchen facilities. The classifi- cation, farmer, exhibits negative relationships with housing con- dition for all of the dependent variables except six rooms or more, exclusive access to kitchen facilities, and telephone available. 201 Comparison Between Estimated Gross and Net Relationships A comparison between the estimated net relationships pre- sented in Table V-8 and the estimated gross relationships presented in Table III-ll reveals that none of the patterns of relationships for comparable housing characteristics are the same. The general pattern of relationships also appears to differ. The estimated gross relationships indicate that the white collar worker has the highest positive relationship to housing condition, farm manager has the second highest, and blue collar worker has third highest. The occu- pational classifications which are associated with lower levels of housing condition exhibited more similarities between their estimated gross and net relationships. Farm laborer was associated with the lowest level of housing condition in both gross and net relationships. In general, the "ranges of effects" for estimated gross relationships are greater than the ”ranges of effects" for estimated net relationships. The only exception occurs with exclusive access to kitchen facilities where the situation is reversed. Type of Tenure The estimated net relationships between the type of tenure variables and measures of housing condition are presented in Table V-9. The tenure category rented (X42) has been omitted from all regression models. Thus the coefficient for this variable is set equal to zero. The tenure category owned (X41) exhibits positive relation- ships with housing condition over the rented category for all measures 202 .NH->H nmnonnn Nu>H moaneh Eonw noxen mnnononmmooo nOnmmonmon onn one mmnnmnonnenon non ooneEnnmo omonh ”oonnom .nn noooooo on oooneoo on xmozH one H->H onnek nn oonnomonm one on» u H» monnenne> .noxen one mnnononmmooo ooneEnnmo onn gong: Eonw mnonmmonwon onn now monnenne> nnoonomoo Anennn onn one nOnnnonoo mnnmnon mo monomeoe omoche omoo.- mnne.m xeozn eeeo. neon. onoonno>< ooonoonon on» eooo. oeno. monnnnnooo ooeoone on ooooo< oonoonoxe on 88. 23. 22333 95.8: e» nooo.- oeeo. onoononoe onoz no ooo n» ooeo. nnon. oeon on omon none nnnoe on Hmhcol Hwho. H030£m .HO £96m d O“ mm000< 0>flW3HUXN m> nooo.- eemo. oonoon ooono nonoz onoo one no: on oeeo.- neoo. oonoonoonno nnnononoonne noz en eono.- moon. ooooe nnnononoonne en nonn. Home. onoz no onoom xne n» noon oooo no ooozo Hex oz x enonnnonou mnnmno: mo monomeoz onnnon mo omxb mnonnnonou mnneso: mo monomeoz one onsnob mo ooze nooznom mannmnonnenom noz ooneEnnmmnn.ou> mamH nmsonnn N->H monnen Eonm noxen mnnononmmooo nOnmmonmon onn one mannmnonnenon non oonennnmo omonh "oonnom OH H on xeozn one n->n onoen nn oonnoeonn one n - n eonnenne> .nn nonneno nn oonneoo .noxen one mnnononmmooo ooneEnnmo onn nuns: Eonm mnonmmonmon onn now moanenne> nnoonomoo xnennp onn one nOnnnonoo mnnmnon mo monomeon omonbe enon. neon. neozn eeoo. oneo. onnenne>< ononnonon on» nooo. nooo.- eonnnnnoen nononnn on eeoooe oonennoxe on nmoo. Nnno. nnononnoe ennneo: en oooo. Homo. enoonnnee onoz no ono n» neoo. meoo. oeon on omon nonn nnnne on Homo. nozonm no nnem e on mmooo< o>nmnnoxm m» eooo.- ooNo. oononn ooono none: onoo one no: on oooo.- eooo. ooneonoenno nnnennnonnne noz en nnoo. ooNo. onooe nnnennnonnne en oeoo. oono. onoz no enoon xne n» eneon e.onx on oneon e.oow on no>oo Henonneonom mox Ho>oq Henonneonom oox oeo: ononomno: mo nonneonom enonnnonou mnnmso: mo monomeoz nonnnonou mnnmno: mo monumeoz one oeo: ononomno: onn mo nonneonom onn nooznom mannmnOnnenom noz ooneEnnmmuu.oHu> mqmH nmsonnn ~->H moanen sonm noxen mnnononmmooo nonmmonuon onn one mannmnonnenon non oonesnnmo omonh .HH HOHQMSU CM vOCMmov ma meZH flaw HI>H QHDQP GM flouflomOHQ 0km .ooo.n» on nenvo now an ooo.n» nenn noneonu oaounn ononomno: o" >0 a n eoneenne> p ”connom .noxen one mnnonUnmmooo ooneanneo onn gonna Eonm mnonmmonuon onn now monpenne> nnoononoo nnennn onn one nonnnonoo mnnmnon mo monnmeoa omonke moon.e monm.on- monm.n nenn.n- onen. neozn omen. momm. - oeeo. mmeo. - memo. onnenne>< ononnonon onn neoo. oeoo. oeoo. omeo. nooo. eonnnnnoen nononnn o OH mmouu< 0>wmflmuxm > eonn. nnen. - moeo. neoo. - mnno. nnononnoe ennneo: en nnmn. oenm. - onno. Hmoo. nnno. enoonnnee onoz no ono nn mmno. neeo. - onmo. nnno. - nmoo. oeon on oeon nonn nnnne on neon. mnon. - oeeo. enno. nomo. nozone no nnee m m OH mmooo< 0>mm3~uxm r moon. nnom. - mnmo. neeo. - ooNo. oonenn ooono o none: onoo one no: n nneo. nmnn. - onno. mono. - eono. ooneonnenno m nnnennnoannm noz > moon. nnno. - oeoo. enno. - neno. onnoe nnnennnonnne en noon. eeme. - onmo. onen. - eooo.- onoz no onoon nne n» poEoonH noeounn nosoonm ononomno: ononomnox finaoononnno ononomno: nnmoononnno mo .moa no .eoo oonnz oox no .eoo eonnnnen non eneoon>nonn ennonenonwnmmmweeo ooxnz omx eonnnaem ovx oonenonno cox . o: m z monxn moonne> mo mononomno: now oeoonn ononomno: nonnnonou unnmnoz mo monomeoz one oaoonn ononomno: nooznom mannmnonnenom noz ooneannmmus.nn-> m4mH nmnonnn N->H moaneh Eonm noxen mnnononmmooo nOnmmonmon onn one mannmnOnneHon non ooneennmo omonh no en xmozH one H->H onpen nn oonnomonm one OH H .HH nonaenu nn oo unnom nnmoo r u > moanenne> .noxen one mnnononmmooo ooneEnnmo onn nuns: Eonm enonmmonmon onn now monnenne> nnoonomoo xnennn onn one nonnnonoo mnnmnon mo monomeoe oeonne eooo. nnen.- neozn eooo.- mnno.- onnenne>< ononnonon onn nmno. nnoo. monnnnnoen nononnn on eeooon o>nonnone on mono. meno.- nnononnoe ennneo: en enno. nmno.- onoonnnee onoz no ono nn emoo. meoo. oeon on omon none nnnne on nono. mono.- nozonm no nnem e on mmooo< o>nmnnoxm m> nnoo.- oono.- oonenn ooono none: onoo one no: on enoo.- oono.- ooneonoenno nnnennnonnne noz mn oeno.- oooo.- onnoe nnnennnonnne n» onno.- emoo. onoz no onoon nne n» omw-W% me zonMWWMMoo me enonnnonou wnnmno: mo monnmeoz nonoonomoo nannnonou mnnmno: mo monnmeoz one nonoonomoo noo3nom mannmnOnneHom noz ooneEnnmmna.NHu> mom ten and a half years). Even so, the effects of education in the second range on housing condition (INDEX) were significantly different from zero at the < .0005 level (positive). It was not surprising to find education having relatively large gross effects on housing condition. Low levels of education are known to be associated with low income levels which in turn are associated with poor housing conditions. However, even with the effects of income and other occupant characteristics removed, edu- cation still has a substantial effect on housing condition. Occupational Classification The occupation of the household head was the third occupant characteristic found to explain a relatively large proportion of the variation in housing condition in both the gross and net relation- ships. Certain classifications were found to have opposite effects depending upon the housing characteristic in question. Farmers are likely to occupy a housing unit with six rooms or more but not likely to enjoy most of the other desirable housing characteristics. Most of the other classifications do not exhibit Opposite relationships. The net relationships reveal that service workers, farmers, laborers, and farm laborers experience the poorest housing conditions. Esti- mates from this same model indicate that farm managers, farm foremen, and white collar workers enjoy the better housing conditions. 225 Size of Place The residence categories, called here "size of place,” exhibited relatively smaller net effects than gross effects on mea— sures of housing condition. As was noted earlier these character- istics exhibited some opposite gross and net relationships. As the population increased the likelihood that the household enjoyed a unit with six or more rooms or exclusive access to kitchen facilities de- creased. For the other measures increasing population has a positive effect on housing condition. The plumbing and heating characteristics of housing are the most strongly affected by these residence categories. Rural farm and rural nonfarm residents are the least likely to enjoy any of the four more desirable types of heating equipment. The estimated net relationships with INDEX also indicate this pattern of relationships. The rural farm residents have the lowest levels of housing condition, followed by rural nonfarm residents. Then the condition of housing increases as the population increases from places of 2,500 population to places of 1,000,000 and more. Residents of urban territories out- side of places experience almost the same levels of housing condition as residents of places that have 1,000,000 and more population. This evidence only adds to the already substantial volume of evidence pointing to the higher incidence and large total amount of poor housing in rural areas and smaller places. According to the Census in 1960, 63.7 percent of all substandard housing units were located outside of Standard Metropolitan Statistical Areas (SMSA), while only 37.6 percent of all housing units fall in this residence 226 category [34]. By 1968, the Current Population Survey indicates that the percentage of all units located outside SMSAs decreased slightly to 36.8 percent while the percentage of all substandard units located outside of SMSAs increased to 67.3 percent. This evidence suggests that government programs to improve housing should be focused in rural areas and smaller places. However, government programs presently are not focused this direction. The United States government response to housing problems began in the mid-1930's with some public housing under public works and related programs. Since that time the annual output of federally assisted housing starts has increased. Through 1969 the number of assisted starts totaled 1,440,300. Thirty-four percent of these were in non-metropolitan areas. Of the 803,700 public housing units only 21 percent were located in non-metrOpolitan areas. Twenty-one and three-tenths percent of the total number of assisted starts were handled through FHA programs. Only 11 percent of these FHA starts were in rural areas. FHA assisted starts totaled 329,300 and 87 percent of these were in non-metropolitan areas [19]. Also presently the two agencies primarily responsible for implementing housing policy miss a significant proportion of the United States population located in smaller places. The Farmer's Home Administration (FmHA) has a legislative mandate to operate in places with 5,500 population or less. The Federal Housing Adminis- tration (FHA) is said to be ineffective in places of less than 25,000 population [19]. According to the 1970 Census 16.9 percent of our population live in places of 5,000 to 25,000 population. 227 This means that significant federal housing programs are not avail- able to almost 16.9 percent of United States citizens. The evidence on housing conditions and government response indicate that national housing policy needs to be directed toward residents of rural areas and smaller places. A Type of Tenure The type of tenure variables were also important in explain- ing the variation in several measures of housing condition. Even with the multicolinearity involved, the dummy variable indicating owner occupancy explained 5 percent of the total variation in INDEX, according to the R2 delete. With each of the measures of housing condition and for both gross and net relationships the effect of owner occupancy relative to renter status was always positive. These results support the past and present housing policy emphasis on home ownership. They also could be used to support an effective home ownership policy for low income families. Race of the Household Head The estimated gross and net effects of the race variables on housing condition add support to previous evidence regarding racial disparities. With each of the measures, household heads who were Negro or had a Spanish surname experienced lower levels of housing condition than white household heads. The finding that the net relationships were also negative indicates that even with the effects of lower educational and income levels removed, household 228 heads who are from a minority background still experience lower levels of housing condition. If these disparities are to be ameliorated, housing policy must be directed disproportionately toward minority groups. Statistical Significance The multiple regression models used to estimate the net relationships between the socio-economic and locational character- istics of the occupants and measures of housing condition, have binary dependent variables. As a consequence statistical tests of the regression coefficients using ordinary least squares estimates of the variances are not valid. The only multiple regression model for which tests using OLS estimates are valid then is the one with INDEX as the dependent variable. A surprisingly large number of the coefficients tested significantly different from zero at < .005 level of significance. Only five of the estimated coefficients were not statistically different from zero at < .05 level of significance. In most cases the relevant statistical test would be a test for the equality of two coefficients. However, the test against zero does indicate that a high percentage of variables exhibit a statistically significant relationship with INDEX. It should also be noted that less than half of the variation (.4488) of INDEX was explained by our independent variables. How- ever, with national, cross-sectional and single household data a large variation within the sample could be expected. 229 Net versus Gross Most of the estimated net relationships with individual desirable housing characteristics differed from the estimated gross relationships only in magnitude. However, for three sets of socio- economic and locational characteristics--sex of household head, metrOpolitan residence in 1955, and dependency--the net relationships were in a different direction than the gross relationships. The changes with the variable female head of household are most notice- able. All of the estimated gross relationships between the presence of a female household head and the desirable housing characteristics are negative. However, seven of the ten estimated net relationships with the binary dependent variables exhibit a positive relationship with housing condition. In most cases the estimated gross relationships have greater "ranges of effects” than the estimated net relationships. The ex- cluded variables, which vary consistently with the explanatory variable which is being studied, cause the range of the estimated gross relationships to be overstated. However, in several cases, which are presented in the summary of Chapter V, the ”range of effects" are greater for the estimated net relationships than for the estimated gross relationships. For these cases the "ranges of effects" for the estimated gross relationships were decreased by the uncontrolled explanatory variables. 230 Further Research Needs The suggestions for further research fall into two cate- gories: (1) those concerning the relationships between socio- economic and locational characteristics of the occupants and housing condition, and (2) those concerning the measurement of housing condition. Measurement of Housing Condition Work presented in Appendix I indicates that present measures of housing condition are probably inadequate for most policy decisions. The measure constructed in this study (INDEX), although an improvement over those presently used, has significant deficiencies. Other mea- sures are needed in order adequately to describe housing conditions and then formulate national housing policy. A limited list of re— search tepics is suggested here: 1. Research is needed to determine the physical characteristics which should be included in a nationally used measure of housing condition. 2. A scale study of satisfaction levels is needed to determine the importance of various physical housing characteristics relative to housing condition. 3. A socially acceptable level of housing condition needs to be determined and a methodology devised to re-estimate this level as social conditions dictate. 231 The second category of research needs concern the relation- ships between socio-economic and locational characteristics of the occupants and housing condition. 1. The relationships examined in this study and others could be used to evaluate the effectiveness of United States housing policies. Specifically housing programs should be examined in the light of their stated objectives, their actual impact, and these studied relationships. Some of the evidence presented in Promises to Keep; Housing Need and Federal Failure in Rural America [19] indi- cate that national policies may be directed away from rather than toward their stated target populations. 2. Work is also needed to examine the administrative frame- work for and the cost of administering housing programs which would meet presently stated goals. A cursory examination of present hous- ing program performance [19] indicates that the costs of meeting stated goals have not been totally reckoned with. Added infor- mation is needed to facilitate bringing funding in line with stated goals. BIBLIOGRAPHY 10. BIBLIOGRAPHY American Public Health Association. Committee on the Hygiene of Housing. An Appraisal Method for Measuring the Quality of Housing:_ Part I. New York: American Public Health Associ- ation, 1945. Anderson, Theodore Wilbur. An Introduction to Multivariate Statistical Analysis. New York: John Wiley and Sons, 1958, chapter 12. Beyer, Glenn H. Housing and Society. New York: Macmillan, 1965. Cooley and Lohnes. Multivariate Procedures for the Behavioral Sciences. New York: John Wiley and Sons, 1962, chapter 3. Edwards, Allen D., and Jones, Dorothy G. Housing in South Carolina: Its Socioeconomic Context. Agricultural Experi— ment Station, Bulletin 511. Clemson, S.C.: Clemson College, April, 1964. Hartmans, Ermond H. "Some Economic and Physical Aspects of Farm Housing and Service Buildings in Selective Areas of Michigan." Unpublished Ph.D. dissertation, Michigan State University, 1950. Hotelling, H. "Relations Between Two Sets of Variates." Biometrika, XXVIII (1936), 321—77. Johnson, Glenn L. "Implications of the IMS for Study of Responses to Price." A Study of Managerial Processes of Midwestern Farmers. Edited by Glenn L. Johnson, Albert N. Halter, Harold R. Jensen, and D. Woods Thomas. Johnston, J. Econometric Methods. New York: McGraw-Hill Book Co., Inc., 1963. Kmenta, Jan. Elements of Econometrics. New York: Macmillan Company, 1971. 232 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 233 Lesgold, Alan M. Analysis of Contipgency Tables: ACT 11. Technical Report No. 14. Computer Institute for Social Science Research. East Lansing, Mich.: Michigan State University, January 12, 1968. Mood, Alexander M., and Graybill, Franklin A. Introduction to the Theory of Statistics. Second Edition. New York: McGraw—Hill Book Co., Inc., 1963, pp. 311—19. President's National Advisory Commission on Rural Poverty. The People Left Behind. Washington, D.C.: Government Printing Office, September, 1967. Rural Poverty in the United States. Washington, D.C.: Government Printing Office, May, 1968. Schaeffer, Annette, and Edwards, Carlton M. A Housing Quality Measuring Scale. Rural Manpower Center Report No. 8. East Lansing, Mich.: Michigan State University, May, 1967. Spurlock, Hughes H. Rural Housipg Conditions in the Arkansas, Missouri, and Oklahoma Ozarks. Agricultural Experiment Station, Bulletin 736. Fayetteville, Ark.: University of Arkansas, December, 1968. Rural Housinnguality in the Ozark Region as Related to Characteristics of Housing Units and Occupants, 1966. Agricultural Experiment Station, Bulletin 758. Fayetteville, Ark.: University of Arkansas, June, 1970. Srikantan, K. S. "Canonical Association Between Nominal Measurements." Journal of the American Statistical Associ- ation, LXV, No. I-492 (March, 1970), 284-92. U.S. Congress. Senate. Committee on Nutrition and Human Needs. Promises to Keep: Housing Need and Federal Failure in Rural America. Washington, D.C.: Government Printing Office, 1971. U.S. Department of Agriculture, Economic Research Service. Quality of Rural and Urban Housing in the Appalachian Region, by Anthony L. Pavlick and Robert I. Coltrane. Agricultural Economic Report No. 52. Washington, D.C.: Government Print- ing Office, April, 1964. Rural Housing in the Northeast Coastal Plain Area of South Carolina, by Robert L. Hurst. Agricultural Economic Report No. 163. Washington, D.C.: Government Printing Office, July, 1969. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 234 Department of Agriculture, Economic Research Service. Rural Housing; Trends and Prospects, by Robert E. Freeman. Agricultural Economic Report No. 193. Washington, D.C.: Government Printing Office, September, 1970. Status of Rural Housing in the United States, by Ronald Bird, Lucia Beverly, and Anne Simmons. Agricultural Economic Report No. 144. Washington, D.C.: Government Printing Office, September, 1968. Department of Commerce. Bureau of the Census. 1960 Censuses of Population and Housing: Procedural History- Washington, D.C.: Government Printing Office, 1966. Evaluation and Research Programs of the U.S. Censuses of Pqpplation and Housing, 1960: Accuracy of Data on Housing Characteristics. Series ER 60, No. 3, 1964. Washington, D.C.: Government Printing Office, 1964. Evaluation and Research Program of the U.S. Censuses of Pqpulation and Housing, 1960: Accuracy of Data on Pqpulation Characteristics as Measured by CPS. Series ER 60, No. 6. Washington, D.C.: Government Printing Office, 1964. Evaluation and Research Program of the U.S. Censuses of Pqpulation and Housing, 1960: Accuracy of Data on Popu- lation Characteristics as Measured by Reinterviews. Series ER 60, No. 4. Washington, D.C.: Government Printing Office, 1964. Evaluation and Research Program of the U.S. Censuses of Population and Housing, 1960: Background, Procedures, and Forms. Series ER 60, No. 1. Washington, D.C.: Govern- ment Printing Office, 1963. Housing Items in the 1960 Census Tract Reports, by Daniel B. Rathbun. Bureau of the Census Working Paper No. 12. Washington, D.C.: Government Printing Office, 1962. Measuring the Quality of Housing, An Appraisal of Census Statistics and Methods. Bureau of the Census Working Paper No. 25. Washington, D.C.: Government Printing Office, 1967. People of Rural America, by Dale E. Hathaway, J. Allan Beegle, and W. Keith Bryant, Census Monograph. Washington, D.C.: Government Printing Office, 1968. 32. 33. 34. 35. 36. 37. 38. Small Areas, pt. 1, U.S. Summary. U.S. Department of Commerce. Bureau of the Census. Pocket Data Book, USA, 1969. Washington, D.C.: Government Print- ing Office, 1969. Procedural Report on the 1960 Censuses of Population and Housing, Bureau of the Census Working Paper No. 16. Washington, D.C.: Government Printing Office, 1963. U.S. Census of Housing) 1960. Vol. 1, States and U.S. Census of Housing, 1960, Final Report PC(l)-1D, Detailed Characteristics, U.S. Summary. U.S. Censuses of Population and Housing) 1960. Description and Technical Documentation. [This contains a description of the data used in this research.] Weisgerber, P. Rural Poverty: The Socioeconomic Characteristics of Low Income Rural Families as Related to Levels of Living. Agricultural Economics Report No. 90, Agricultural Economics Department. East Lansing, Mich.: Michigan State University, March, 1968. Williams, A. V. Canonical Analysis: CANON. Technical Report 32, Computer Institute for Social Science Research. East Lansing, Mich.: Michigan State University, October 28, 1965. APPENDICES APPENDIX I REPRESENTATIVENESS OF STRUCTURAL CONDITION APPENDIX I REPRESENTATIVENESS OF STRUCTURAL CONDITION Examining the Census measure of structural condition for representativeness of general housing condition was not part of the initial research proposal. However, as the work progressed, it be— came apparent that this should be included as a secondary objective for several reasons. First, evidence was discovered which indicated that structural condition may not be representative of general housing condition. This evidence will be presented later on in this appendix. Secondly, the literature exhibits an acceptance of the assumption that structural condition is representative of general housing condition. The Assumed Hypotheses An obvious indication of this belief is the common reference to structural condition as a measure of housing "quality." The U.S. Bureau of the Census, in an attempt to rate the quality of housing in 1960, used three classifications of housing quality--sound, deteriorating, and dilapidated [23, p. 4]. Also, it is suggested that housing units that are sound and have complete plumbing facilities have other good housing qualities. 236 237 The Bureau of the Census has adopted a combination of sound- ness of structure and completeness of plumbing facilities as a partial standard for measuring quality. Such factors ad adequate lighting and ventilation, and the neighborhood also are recog- nized as quality factors, but the Bureau points out that these qualities are difficult to measure in a broad Census enumeration. Also, these qualities are generally found packaged-in with houses that are sound and have complete plumbing [16, p. 23]. A related assumption is that various other characteristics of housing are, in fact, representative of general housing condition. For example, if age of housing or plumbing facilities are representa- tive of general housing condition which would include structural condition, then structural condition should represent plumbing facil- ities, the age of housing, and general housing condition. Spurlock states some of these assumptions. the To obtain an operational indicator of adequate housing, the 1,413 respondents were grouped into three categories. The category with complete plumbing includes all housing units in the sample with the following: hot and cold running water, inside; a flush toilet, inside; a bathtub or shower; a commercial water supply or drilled well; and access to a public sewer or septic tank. Such housing units were designated as adequate. It was assumed that such housing would generally be structurally sound and adequate in other quality aSpects, though there are undoubtedly exceptions. In this report, the terms with complete plumbing, with partial plumbing, and with no plumbing are used interchangeably with adequate housing, partially adequate housing, and inadequate housing, respectively [17, p. 6]. The age of housing may be indicative of its quality. As a general rule, older houses have fewer modern features, are more likely to be dilapidated, and are often in need of extensive remodeling or repair [16, p. 13]. Bird, Beverly and Simmons also state the assumption regarding age of housing: An inventory on the age of housing units can be a rough measurement of the adequacy of housing and of trends in housing construction [23, p. 3]. 238 Two other assumptions for which there is some empirical evi- dence tend to support the general assumption that structural condition is representative of general housing condition. One is that an index of general housing condition including structural condition is in- sensitive to weight changes. And the second is that individual measures of housing condition are highly positively correlated. Weisgerber, when constructing an index of housing condition from 17 separate measures, stated that: In trying to arrive at a satisfactory weighting system for combining the various factors into a single index, several variations based on relative factor importance were tested. The net rating for each dwelling was found not to change a great deal as several plausible weighting systems were tried [37, p. 101]. The indicated insensitivity to changing weights would sug- gest that the included measures are positively correlated. Two other studies indicate the existence of a high positive correlation between individual measures of housing condition [6, 15]. It is not my contention that any of the individual studies cited argues strongly for the assumption that structural condition is representative of general housing condition but that a review of these works can lead to the conclusions that: (l) the "important" measures of housing condition are highly-positively correlated, and (2) some of those mentioned including structural condition are repre- sentative of general housing condition. The assumption to be examined here--structural condition as measured by the Census represents general housing condition-~is diffi— cult to test using Census data for several reasons. First, only a small number of other measures are included. As a consequence, 239 structural condition could be highly correlated with each one and still not be representative. A second difficulty is a measurement problem. What level of correlation must structural condition have with each other measure or combination of measures in order to be either highly correlated or, the more basic question, to be repre- sentative? Lacking definitive solutions to these difficulties we will provide information on but not test the basic question specifi- cally. Assumptions Several assumptions are presented to establish the basis for examining representativeness. 1. Housing condition is a multi-dimensional concept including more than just structural condition. This has been brought out clearly in our discussion of theoretical considerations in Chapter 2. 2. Other measures of housing conditions included in Census data are a part of general housing condition. An examination of the Census measures (Table A-I-l) will reveal that they are similar to some of the items included in other measures-- Schaeffer and Edwards [15] and the APHA method [3]. 3. Each of the other Census measures and the measure of struc- tural condition can be ranked ordinally with respect to their relationship to general housing condition. This has been done in Table A-I-l. 240 TABLE A—I-l.--Parameter Estimates from Canonical Correlation: Structural Condition = f (Other Measures of Housing Condition) Variables Includedb Parameter Estimates a. and b. J J 1. Structural Condition X Sound 1.1631 X Deteriorating - .9599 X3 Dilapidated —2.5474 2. Telephone Y1 Telephone Available .5089 No Telephone 0a 3. Kitchen Facilities Direct Access, Exclusive Use 0a Y Direct Shared Access or No Equipment .1021 Y4 Access Through Another Unit - .3225 4. Water Supply Hot and Cold Piped In 03 Y5 Cold Piped Inside -l.l608 Y6 Water Piped Outside —2.094l Y7 No Piped Water -l.7850 5. Year Built Y8 1959 through March 1960 .7266 Y9 1955 through 1958 .6734 Y10 1950 through 1954 .6202 Y11 1940 through 1949 .4538 Y12 1930 through 1939 .2789 1929 or earlier 08 6. Heating Equipment Y13 Built-in Electric Units .3649 Y14 Steam or Hot Water .6023 Y15 Warm Air Furnace .4669 Y16 Floor, Wall or Pipeless Furnace a.‘4684 Other Means, With Flue 0 Y17 Other Means, No Flue - .0769 Y Not Heated - .3771 18 TABLE A-I-l.--Continued 241 Variables Includedb Parameter Estimates 3. and b. J J 7. Number of Rooms Y19 10 or More Rooms .1651 Y20 9 Rooms .1579 Y21 8 Rooms .1489 Y22 7 Rooms .1759 Y23 6 Rooms .1625 Y24 5 Rooms .0931 Y25 4 Rooms .0022 Y26 3 Rooms - .0591 Y27 2 Rooms - .1904 1 Room 0a 9. Access to a Flush Toilet Exclusive Shared None 3. Access to Exclusive Y .6758 Y - .7630 0a 28 31 a Bath or Shower Shared Y29 .8523 Y32 - .1117 Y34 —l.8506 None Y30 -.l46l Y33 -l.2430 Y35 .2426 a . . This variable was bAll variables are holds and zero otherwise. Source: Census tapes from of POpulation and omitted to avoid singularity. dichotomous equalling one if the condition the one-in—a-thousand sample, 1960 Censuses Housing, 25 percent sample portion [36]. 242 It is assumed that having a telephone available is a higher level of condition than no telephone. With respect to kitchan facili- ties, it is believed that having direct access with exclusive use is the highest level of condition with direct shared access or no equipment being the next level and access to facilities through another unit being the lowest level of condition. For the measure called water supply hot and cold water piped inside is designated the highest level of condition, cold water piped inside, next, fol- lowed by water piped outside and the lowest level being no piped water. For the measure, year built, it is assumed that the newer the higher the condition level. The highest level of heating equipment is assumed to be built-in electric units; the next level, steam or hot water; followed by warm air furnace; then by floor, wall or pipe- less furnace; next, other means with flue; then other means, no flue; and the lowest level of condition, not heated. It is assumed with the next measure that the more rooms in the housing unit, the higher the condition level. The next two measures of housing condition, access to a flush toilet and access to a bath or shower, each have three levels of condition going from highest to lowest, exclusive, shared and none respectively. Also these two measures have been com— bined to make nine relative condition levels. Exclusive use of a bath or shower and flush toilet is assumed to be a higher level than shared use of both which is higher than no access to either one. Condition levels are also ranked from highest to lowest as the access to one item is held constant while the other is varied from exclusive to none . 243 4. Structural condition should vary consistently with the ordinal ranking of most of these other housing condition measures or it does not generally reflect housing condition. Canonical Correlation It is believed as stated in Assumption 4 above that struc- tural condition should exhibit a positive relationship with other measures of housing condition if it is to be representative of general housing condition. Therefore, the three research methods employed here examine the data for a positive relationship. The first, canonical correlation, is used to estimate net relationships between structural condition and other measures of housing condition. Each of the levels of housing condition is represented by a binary variable as presented in Table A-I-l. For example: X1 = 1 if the unit is sound 0 otherwise X2 = 1 if the unit is deteriorating 0 otherwise X = 1 if the unit is dilapidated 0 otherwise The Yi (i = 1, 2, . . ., 35) are also binary variables equalling 1 if the condition holds and zero otherwise. These binary variables are then combined linearly. The 1th observation would look like this: a1x11 l a2x12 * a3X13 = x1 b1Y11 + bZYiZ 1 * bssyiss ’ Y1 Where: Xij and Yij are the binary variables presented in Table A—I-l, bj and aj are the coefficients to be estimated, and Xi and Yi are the linear combinations of the X's and Y's, respectively, or canonical variates. aj and bj are estimated such that the correlation between Xi and Y1 is maximized. Canonical correlation was used for several reasons. First, it can provide estimates of the unique set of net relationships between two sets of variables which provides maximum correlation. Secondly, it allows for all variables to be binary. And lastly, it allows for a stochastic component in both sets of variables. A fur— ther discussion of this model and its characteristics can be found in Appendix III. Empirical Results The results of the canonical correlation analysis are pre- sented in Tables A-I-l and A-I-2. Only the parameter estimates for the first canonical correlation coefficient are presented even though all were significant at <.005 level of significance as can be seen in Table A—I-Z. This was done because we are interested in that set of coefficients which yields the maximum correlation between structural condition and other measures of housing condition. 24S .Homg nOnnnon onenem nnoo -non mN .mnnmno: one nonnennaoo mo momnmnou ooon .oHQEem onemnonn-e-nn-ono onn eonm women mnmnou “oonnom .nmon nmnnm onn nn ooneonMnnMnm mo Ho>on moo.v ne oonoomon on or monnnnn ooneonmnnwnm mo Ho>on moo.v ne oonoomon mnnon oz nenn oonno moononomon one HHH xnononn< eonm onon Hans nooeon onn nan .nmon mnnn now oonnenonoo non me: onenom-nno one o .nonnenonnoo neonnoneo no nonnoom onn .HHH xnonoao< nn oonnenexo one onon oomn mnmon neonnmnnenm onk n o .o o eomn. u m o noom. u o n meem. u n n .nno>nnoooeon . . . we no no mnonnenonnoo neonnoneo neonnenm one nmomnen nxon nmomnen onn one m one . m. m e moo.v mm nm~.mom .onoN on nenoo on mom nnnmnonnenon one no: .m moo.v oo non.ooon .onoe on nenoo one mon one mom mnnnmnonnenon oonnnnoo onb no: .N . . mo omoo v onoe on Hesoo one m one o N m nHU Q . . m m nnmnonnenon oonnneoo onk .o: H o oonoonon on Ho>oo Eoooonm mo onenomnnnu momonnoaxx nnnz moonmoo ooneonMnnmnm nonnnonou mnnmno: mo monomeoz nonno one nonnnonoo Hennnonnnm nooznom mnnnmnonnenom neonno now .eeneon--.m-nu< mnm ' ' Ho. Bi Bj (1 # j) The test statistic has the students t distribution if the null hypothesis is true. 8. - 8. _._—;L1 Nt [9. P- 372] SB. _ B n-k kin 275 Where: (1) Bi and Bj are ordinary least squares estimates of 8i and Bj, reSpectively. (2) Sgi - 8j is the estimated standard deviation of the difference between the two parameters. The last hypothesis is that one Bj is equal to zero. Hozfij = 0. The test statistic has the students t distribution if the null hypothesis is true. If“ Nt (D ID) n~k [9, p. 118] 3' Where: (1) Bj is an ordinary least squares estimate of Bj' (2) $8 is the estimated standard deviation of Bj. 5 Canonical Correlation2 The third statistical tool used is canonical correlation. It has some similarities with regression analysis. Multiple re- gression can be used to estimate the net relationships between a set of variables, the predetermined variables, and a single variable, the endogenous variable. Canonical correlation, on the other hand, can be used to estimate the net relationships between two sets of variables. Also, canonical correlation assumes that both sets of variables have a stochastic component. In the case where one set of variables is reduced to only one variable, canonical correlation 2The computer program used was made available through the Computer Institute for Social Science Research at Michigan State University. The program, described in the Institute's Technical Report 32, Canonical Analysis: CANON [38], is also available in Multivariate Procedures for the Behavioral Sciences by W. W. Cooley and P. R. Lohnes [4]. 276 reduces to multiple regression with stochastic predetermined vari- ables. Cooley and Lohnes indicate that: The interrelations between two sets of measurements made on the same subjects can be studied by canonical-correlation methods. As deve10ped by Hotelling (1935, 1936), the canonical correlation is the maximum correlation between linear functions of the two sets of variables. Several linear combinations of the two sets are frequently possible. Each pair of functions is so determined as to maximize the correlation between the new pair of canonical variates, subject to the restriction that they be independent of previously derived linear combina- tions [4, p. 35]. This tool is used in Appendix I to examine the net relation- ships between the set of binary variables representing the struc— tural condition of housing and the set of binary variables represent- ing the other measures of housing condition. The model is described as: Xi = alxil + azxi2 + 83x13 Yi = blxil + b2x12 + . . . + b35x135 Where: (1) Xi = a linear combination of the xij for the ith observation and is called a canonical variate. (2) Y1 = a linear combination of the Yij for the jth observation and is called a canonical variate. (3) xij and Yij are binary variables describing the ith observation. (4) aj and bj are the coefficients used in the linear combination of the X's and Y's, respectively. A set of coefficients, aj's and bj's, are estimated such that the correlation between Xi and Y1 is maximized. Since the smaller set, the X's, contains three variables, three independent sets of coefficients can be estimated. That is, each pair of 277 canonical variates is uncorrelated with the other pairs of canonical variates and has maximum correlation [2, p. 295]. In the model used in this research, we have three canonical correlation co- effici n s--R R ' ' e t cl’ c2’ and Rc3 With RC1 being the largest, Rc next 2 3 the smallest. It should be noted that in the case dis- and Rc cussed above where canonical correlation reduces to multiple re- gression, the one canonical correlation coefficient is the multiple correlation coefficient [2, p. 298]. The tests of Rcl’ Rc2 and Rc3 are nested sequential tests using a statistic with a chi-square distribution [18; 4, p. 37]. The first null hypothesis is He: The two sets of variables are unrelated. If this is rejected, then the second null hypothesis is He: With the effects of the largest canonical correlation coefficient removed the two sets of variables are unrelated. If this one is rejected, then the smallest canonical correlation coefficient is tested with the effects of the larger ones removed. For addition discussion of and references to this statistical tool, see Anderson [2], Cooley and Lohnes [4], and Srikantan [18]. APPENDIX IV ACCURACY OF MEASURES OF HOUSING CONDITION APPENDIX IV ACCURACY OF MEASURES OF HOUSING CONDITION This appendix is included to compare the accuracy of the Census measure of structural condition to other measures of housing condition. Some information is also included relative to sources of information on the accuracy of data pertaining to the socio-economic and locational characteristics of households. However, for pur- poses of this study the data on the characteristics of households are assumed to be measured without error. Measurement Error The 1960 Censuses of Population and Housing contain measure- ment errors from several sources. . the missing of people by enumerators will result in undercounts, personal characteristics may be erroneously reported, people fail to report some of the information requested of them and adjustments for these persons may introduce errors, and so forth [27, p. l]. A number of studies have been conducted by and for the Bureau of the Census to determine the extent of such errors. Of these studies, one, Evaluation and Research Program of the U.S. Censuses of Population and Housing, 1960: Accuracy of Data on Housing Characteristics [25] (referred to as CES, Content Evaluation Study) 278 V I __.i 279 will be used to describe some of the housing condition measures included in the Census. Two other studies from the Evaluation and Research PrOgram of the U.S. Censuses of Population and Housing, 1960 may be of interest to the reader who would like to examine the accuracy of population characteristics: Accuracy of Data on Popula- tion Characteristics As Measured by CPS [26] and Accuracy of Data on Population Characteristics As Measured by Reinterviews [27]. Content Error Vs. Coverage Error These three studies pertain to "content error” rather than the "coverage error." I am assuming that the "coverage error" causes some undercounting of households having low housing condition. Where this is true there will be underestimates of the gross re- lationships between socio-economic and locational characteristics of the households and measures of housing condition. Coverage error, however, should not bias our estimates of net relationships between socio-economic and locational characteristics of households and measures of housing condition. It would only result in fewer ob- servations among the groups undercounted. We are ignoring the bias created by omitted information. But the "content error" pertains to the accuracy of the individual record and is of concern here. The U.S. Bureau of the Census uses a number of special statistics to analyze the reinterview information. The following is a description of those statistics, how they are constructed and what they mean. 280 Census Measures of Accuracy Indexes of Response Variance and Bias The response errors of a particular census or sample survey result from the joint effects of response bias and response variance. Measures of these two items can therefore be used as indexes of the accuracy of the data. A brief description of reSponse bias is that it represents systematic errors in re- porting data, or the effect of types of errors that are con- sistent in direction and that would be consistent if it were possible to do independent repetitions of the survey under the same general conditions. Response variance, on the other hand, can be categorized as the effect of errors which tend to cancel out when a large number of observations are made. The para- graphs which follow give a more complete description of these terms. For a fuller description, see the report Series ER 60, No. 1, Evaluation and Research Program of the U.S. Censuses of Population and Housing, 1960: Background, Procedures, and Forms and the references in the bibliography of that report. Under certain fairly general survey conditions, matching information from two sources for identical persons can provide estimates of response variance, and to the extent that one of these sources is based on more adequate measurement methods and is acceptable as a standard, it can also provide estimates of bias. Various measures of response variance and bias can then be constructed from the results of this kind of match. The CES, compared with the census, gives two measurements for each person reinterviewed for selected items of information and roughly satisfies the conditions given above. A group of such measures, which appear to be useful for analytic purposes, have been computed for each characteristic Studied and are shown in Table A-IV-l. TABLE A-IV-l.--General Representation of Results of Original and Reinterview Surveys of Identical Persons Results of Census Results of the CBS 1 0 Total 1 a b a+b 0 c d c+d Total a+c b+d n = a+b+c+d ‘2. h .4- 281 Table A-IV—l illustrates the results of the comparison of the census with the CES where the value 1 is assigned to a person classified as having some specified characteristic and the value 0 otherwise. (Persons who have no response in either interview for the characteristic being studied are excluded.) Table A-IV—l shows that "a" of the persons were classified as having the specified characteristic in both the census and CES, "a+c” were classified as having the characteristic in the census, and "a+b" were classified as having the characteristic in the b+c (n-l)se (c-b) _ - + n n n2*’ 1. Gross difference rate: g = When n is large, the first component of the gross difference - rate is approximately equal to the simple reSponse variance of “*5 the census statistic when the difference between the CES and the census is used as a measure of the bias. The second com- ponent is the square of the estimated bias of the census statistic. If the bias is small, the gross difference rate can be used as a measure of the simple response variance of the response differences. It can be shown that if the census and a second survey were independently conducted under the same general conditions, the simple response variance of the response difference as developed above would be twice the simple response variance of the census (or of the second procedure). Therefore, under these conditions g/2 would be an approximate measure of the response variance of the census, and is in fact the measure used in this report. However, the CES was not conducted independently. As pointed out earlier differences between information reported in the census and the reinterview were reconciled. This would imply that the measurement g/2 tends to be an underestimate of the variance of the census. i..._2 -.-——s—— 2. Index of inconsistency: 2pq p1q1+P2q2 This index shows the ratio of the simple response variance g/2, to pq where p is the average proportion in the census and CES having the specified characteristic. If the CES is viewed as being a repetition of the census, then pg can be estimated by 1/ P q +P Q'- 1 1 2 2 = Sflfisl- is the proportion of matched persons 2 p1 l/Under other conditions (for example, where there is knowledge that the reinterview survey is subject to much less re— sponse variability than the census and it is desired to compare the quality of two censuses) it would be more appropriate to use a different estimate of pq. In the example mentioned, the comparison 282 in the CES sample having a specified characteristic in the census, p2 = 12%22_ is the proportion of matched persons in the CES sample having that same characteristic in the CES, q = l - p (b+d) (c+d) 1 1 q]. = 1 - p1 = T— and q2 = 1 - p2 = T . Therefore, I is estimated in the following way: f = Cb+C)/n a+c b+d + a+b c+d n n“ n n A simple interpretation of I is as follows: _fl Assume that a sample of n elements is drawn with equal probability and with replacement. Also, assume that the between element covariance of response deviations is zero--that is, that the quality of response of one person is independent of the quality of reSponse for any other person. Then, for a sample of one element, the total variance can be expressed as the binomial variance, pq. The total variance is. then, the sum of the simple response variance and the "pure" sampling variance. Therefbre, the simple response variance is equal to or less than pq. As stated above, g/2 is an estimate of the simple response variance. As the measurement of the specified characteristic becomes less reliable, but remains unbiased, the simple response variance increases and the sampling variance decreases. When the measure— ment process becomes equivalent to tossing the same coin for each element (0