characteristics of the study area and of the data · 2020. 3. 20. · table 4-1 characteristics of...

This PDF is a selection from an out-of-print volume from the NationalBureau of Economic Research

Volume Title: Housing Markets and Racial Discrimination: A MicroeconomicAnalysis

Volume Author/Editor: John F. Kain and John M. Quigley

Volume Publisher: NBER

Volume ISBN: 0-870-14270-4

Volume URL: http://www.nber.org/books/kain75-1

Publication Date: 1975

Chapter Title: Characteristics of the Study Area and of the Data

Chapter Author: John F. Kain, John M. Quigley

Chapter URL: http://www.nber.org/chapters/c3714

Chapter pages in book: (p. 92 - 117)

4

Characteristics of the Study Area andof the Data

INTRODUCTION

The analyses presented in this book are based primarily on anextensive body of data collected in 1967, which describes samples of1,583 units and their occupants in the city of St. Louis, and 206 dwellingunits and their occupants in St. Louis County. The choice of the issuesto be addressed, and the strengths and weaknesses of the methodologyemployed, are intimately related to the nature of the available data andto the composition of the sample. For example, a novel feature of theresearch is an unusually elaborate analysis of the characteristics ofhousing bundles. Most empirical analyses of housing markets have beenbased either on census tract or other aggregate statistics or on samplesof microdata, such as the census one-in-one-thousand sample, whichprovide some information on individual housing units and householdsbut include no information on the neighborhood environment. Thesample used in this study provides extensive information on all threedimensions: household characteristics, dwelling units, and the broaderneighborhood and community environment. The surveys and the tech-niques used to coilect these data may themselves be useful methodologi-cal contributions. Therefore, in this chapter we shall describe thesurveys in some detail.

The extensive analyses of the effects of housing-market discrimina-tion presented in this book are made possible by a systematic oversam-pling of black households. In 1960, blacks accounted for 16 percent of allhouseholds in the St. Louis metropolitan area and 19 percent of thecombined population of St. Louis County and city.' Black householdscomprise 34 percent of the samples used for the analyses reported here.

'U.S. Bureau of the Census, U.S. Censuses of Population and Housing: 1970,"Census Tracts," St. Louis (GPO, 1971), Table P-i.

92

Characteristics of the Study Area and of the Data 93

In contrast, the samples of high-income and suburban neighbor-hoods are far smaller than would be desirable. To some extent, therelatively small samples of high-income households result from thedecision to obtain relatively large samples of black households. How-ever, the suburban sample is considerably smaller than it would havebeen if the analysis of metropolitan housing had been the only considera-tion in sample design. This was not the case. The data used in thisresearch were collected initially for more limited descriptive and analyticpurposes by the St. Louis Community Renewal Program (CRP). Thepriorities of that program dictated a heavy sampling of target neighbor-hoods. Moreover, program restrictions prohibited the CRP from inter-viewing households beyond the boundaries of the city. As has beennoted previously, the small, but exceedingly valuable, sample of house-holds for St. Louis County was financed from other sources.

There are many other instances where the design of the researchthat we undertook was influenced, made possible, or compromised bythe characteristics of the survey instruments used or the composition ofthe sample. For example, the effect of workplace location on housingchoices, which figures prominently in the theoretical discussion pre-sented in Chapter 2, is not evaluated in subsequent empirical analysesbecause of the limited geographical coverage of the sample and thesystemic oversampling of unemployed or marginally employed house-holds.

To understand these issues, it is necessary to describe the data-collection procedures, the sampling methods, and the characteristics ofSt. Louis itself. This chapter is designed to provide this background.First, we present brief descriptions of the St. Louis metropolitan area, ofits population, and of its housing market. These descriptions are fol-lowed by a presentation of the several survey instruments used to collectinformation on sample households, dwelling units, and neighborhoodsand of the measurements drawn from these surveys. The chapter endswith a discussion of the sampling design and a brief comment on alterna.tive methods of weighting sample observations.

CHARACTERISTICS OF THE STUDY AREA

In 1970, nearly two and one-half million persons lived in the St.Louis metropolitan area. Nearly two-thirds of these lived in the city ofSt. Louis and in adjoining St. Louis County. (St. Louis city and St.Louis County, along with East St. Louis, Illinois, and the balance of St.Clair County, Illinois, form the urbanized core of the metropolitan area.)East St. Louis and St. Clair County accounted for an additional 12

94 HOUSING MARKETS AND RACIAL DISCRIMINATION

percent of the metropolitan area population in 1970. The remainder ofthe metropolitan area, shown in Figure 4-1 (St. Charles, Jefferson, andFranidin Counties in Missouri, and Madison County in Illinois),accounts for only 22 percent of the metropolitan population and is onlyloosely linked to the central portions of the region. Housing located inEast St. Louis is more competitive with housing located in St. Louis cityand County, but the interrelationships are not so great that an analysis of

/

p

Source: U.S. Bureau of the Census, U.S. Census ofHousing: 1960, Vol. II, Mefropoliton Housing. Part 5,Newport News-Hampton.Santa Barbara StandardMetropolitan Statistical Areas. Washington, D.C.: U.S.

Printing 1963, p. 1

ST. CHARLES COUNTY

-y(ST. LOUIS COUNTY

Li.MADISON COUNTY

N/ JEFFERSON COUNTY

"UST. CLAIR COUNTY

7_J

— — — — — STATE LINE— - — — COUNTY UNE

*

-

0 10 20 MILES

N

FIGURE 4-1St. Louis Standard Metropolitan Statistical Area


the St. Louis housing market which excludes East St. Louis is seriouslydeficient.

As the data in Table 4-1 illustrate, the city of St. Louis exhibits thecharacteristics common to old U.S. cities, albeit in a somewhat extremeform, while adjoining St. Louis County conforms rather well to thesuburban stereotype.2 Thus, in 1970, 41 percent of St. Louis city'spopulation was black, but blacks comprised only 5 percent of St. LouisCounty's population. Moreover, 63 percent of city residents were rent-ers as contrasted with only 26 percent of St. Louis County residents.Within the renter and owner markets, the median value of single-familyunits in the central city was only 70 percent as great as the comparablevalue for the suburbs; the median rent of city units was only 54 percentas large as that of suburban units.

East St. Louis, Illinois, is even blacker and poorer than St. Louis,Missouri. In 1970, 69 percent of its population was black; the medianvalue of its single-family homes was only $8,800; and the median rent ofits rental units was only $63 per month.

St. Louis exhibits still other characteristics that make it an arche-typal central city. Between 1960 and 1970, the population of St. Louisdeclined by 127,790, or 19 percent, the largest percentage decline experi-enced by any U.S. central city during the decade. From Figure 4-2, it isapparent that these declines are the continuation of a trend which wasevident as early as 1940. The economic expansion and controls associ-ated with the Second World War continued to produce some growth inthe decade 1940—50, but the experience, at least since 1950, has been oneof rapid decline in central-city population. The central city's whitepopulation declined by even more, but part of this decline was offset byan increase of nearly 40,000 in the black population. Suburban St. LouisCounty, by contrast, grew steadily from the end of the First World Warto the end of the Second World War and has grown explosively sincethen. Between 1960 and 1970, suburban St. Louis County grew by248,000, an increase of 218,000 whites and 26,000 blacks. These differ-ences in the growth of black and white populations in the central city andin suburban St. Louis County are clearly shown in Figure 4-3.

As is evident from Figure 4-4, segregation is intense within thecentral city, with black households residing in a few well-defined neigh-borhoods. Close to 98 percent of the city's black population resides inthree community areas: the Model Cities neighborhoods (Yeatman,

2Among central cities of large SMSA's (over 500,000 population), only eight had alarger proportion of black residents in 1970, only six had a more adverse ratio of central-cityto SMSA poverty (percent of St. Louis city population in poverty in 1970/percent of St.Louis SMSA population in poverty 1970 1.9), and none had a larger decline in populationbetween 1960 and 1970.

TA

BLE

4-1

Cha

ract

eris

tics

of M

ajor

Sub

divi

sion

s of

the

St.

Loui

s M

etro

polit

an A

rea:

197

0

%

Subd

ivis

ion

Popu

latio

n%

Bla

ckD

wel

ling

Uni

tsO

wne

r-O

ccup

ied

Med

ian

Val

ueM

edia

nR

ent

Med

ian

Num

ber

of R

oom

s

SMSA

2,36

3,01

716

.078

5,49

860

.5$1

6,30

0$

794.

7St

. Lou

is C

ity62

2,23

640

.923

8,48

536

.613

,200

684.

0St

. Lou

is C

ount

y195

1,35

34.

829

1,65

073

.918

,800

125

5.2

Eas

t St.

Lou

is69

,996

69.1

23,6

1345

.58,

800

634.

3B

alan

ce o

f St

. Cla

ir C

ount

y,Ii

i.21

5,18

07.

067

,741

69.8

13,6

0088

4.6

St. C

harl

es C

ount

y, M

o.92

,954

1.2

28,1

1971

.317

,500

985.

0Je

ffer

son

Cou

nty,

Mo.

105,

248

0.8

33,6

7771

.314

,600

824.

7.

Mad

ison

Cou

nty,

III

.25

0,93

45.

282

,286

70.0

13,9

0076

4.6

Fran

klin

Cou

nty,

Mo.

55,1

161.

119

,927

65.5

13,3

0062

4.7

SOU

RC

E: U

.S. B

urea

u of

the

Cen

sus,

U.S

. Cen

suse

s of

Pop

ulat

ion

Hou

sing

: 197

0, "

Cen

sus

Tra

cts,

" St

. Lou

is (

GPO

,19

70),

Tab

les

P-i,

H-i

.'E

xclu

ding

St.

Lou

is C

ity.

CD a)


'40FIGURE 4-2

Population of the City of St. Louis, St. Louis County, and the Balance of theSt. L0uisSMSA(asDefined in 1970)from 1910to1970[Source: U.S. Bureauof the Census, Fourteenth Census of the United States, 1920. Population,Vol. III, "Characteristics of the Population by States" (GPO), Missouri,Table 9, and Illinois, Table 9; Sixteenth Census of the United States, 1940:Population, "Characteristics of the Population by States" (GPO), Missouri,Table 22, and Illinois, Table 22; Census of Population: 1950, Vol. II,

"Characteristics of the Population" (GPO), Pt. 13, Table 42, and Pt. 25,Table 42: Census of Population: 1960, Vol. I, "Characteristics of the Popula-tion" (GPO), Pt. 15, Table 28, and Pt. 27, Table 28; Censuses of Population

and Housing: 1970, "Census Tracts" (GPO), St. Louis, Table P-i.]

Murphy-Blaxir, Carr-Central, Montgomery—Hyde Park), the near SouthSide, and the West End. The Taeuber and Taeuber index of racialsegregation (discussed in Chapter 3) was 90.5 for St. Louis in Theexpected index of racial residential segregation computed by Taeuber

3Karl E. Taeuber and Alma F. Taeuber, Negroes in Cities: Residential Segregationand Neighborhood Change (Chicago: Aldine Publishing Co., 1965), p. 400.


FIGURE 4-3Black and White Population of St. Louis County and the City of St. Louisfrom 1910 to 1970 [Source: U.S. Bureau of the Census, Fourteenth Censusof the United States, 1920: Population, Vol III, "Characteristics of the Popu-lation by States" (GPO), Missouri, Table 9, and Illinois, Table 9; SixteenthCensus of the United States, 1940: Pdpulation, "Characteristics of thePopulation by States" (GPO), Missouri, Table 22, and Illinois, Table 22;Census of Population: 1950, Vol. II, "Characteristics of the Population"(GPO), Pt. 13, Table 42, and Pt. 25, Table 42; Census of Population: 1960,Vol. I, "Characteristics of the Population" (GPO), Pt. 15, Table 28, and Pt.27, Table 28; Censuses of Population and Housing: 1970, "Census Tracts"

(GPO), St. Louis, Table P-i.]

and Taeuber, based on indirect standardization for tenure, value, andrent of occupied dwelling units, was only 12.4 in the same year.4Therefore, based on these comparisons, in 1960 segregation of the raceswithin the central city was roughly seven times as great as that impliedby tenure choices and housing expense.

4lbid., p. 85.

ition (thousands)

99

FIGURE 4-4Percent Black by Census Tract in 1970 for St. Louis City and County[Source: Hugh 0. Nourse and Donald Phares, "The Filtering Process:Aging or Arbitrage" (Columbia, Mo.: University of Missouri, 1972), pro-

cessed, p. 9.]

Characteristics of thi? Study Area and of the Data

% 1970 1960

ST. LOUIS CITY BOUNDARY

* 1910

o 1960.,,,,l


THE SURVEY INSTRUMENTS

An important feature of the research presented in this book is itsextensive analyses of housing attributes. The basis of these analyses is adetailed description of each sample housing bundle. Although thesedescriptions are still far from ideal for analyzing the behavior of urbanhousing markets, they are unprecedented in their detail and complete-ness.

The data, collected originally by the St. Louis Community RenewalProgram, were used by the St. Louis CRP for several purposes, one ofthe most important of which was the development of a model to projectthe extent of residential blight for each block and census tract of St.Louis. Estimation of this model required detailed data on the quality andcondition of individual dwelling units, blocks, and neighborhoods; thesedata form the basis of the analyses presented

A home-interview survey of 1,583 dwelling units in the city of St.Louis and 206 dwelling units in the St. Louis County suburbs is the mostimportant data source for the subsequent analysis. The city dwelling-unitsample was a stratified random sample in which dwelling units in low-income neighborhoods were sampled at higher rates than dwelling unitsin high-income neighborhoods. St. Louis County dwelling units wereobtained by a stratified random sampling procedure of dwelling unitslocated there.

Nonresponses, refusals, vacancies, demolitions, and language diffi-culties reduced the original sample of 1,789 dwelling units to 1,273 usablehome interviews, of which 1,186 households were in the private housingmarket. Vacancy is the largest single reason for failing to obtain a homeinterview (192, or approximately 11 percent, of the units selected werevacant). Another 14 dweffing units had been demolished and had not asyet been removed from the current City Planning Commission mapsused in drawing the sample. The number of incomplete interviews byreason is enumerated in Table 4-2.

Of the 1,273 sample dwelling units producing usable home inter-views, 658 were renter-occupied (including 50 renters in public housing)and 578 were owner-occupied. An additional 37 units were occupied bytenants who, for a variety of reasons, did not pay rent.

The 87 units which were not in the private housing market (50public-housing units and 37 tenants who did not pay rent) were removed

5The blight model was developed by the authors as consultants to Alan M. Voorheesand Associates, Inc. It is described in Alan M. Voorhees and Associates, Inc., TechnicalReport on a Residential Blight Analysis for St. Louis, Mo. (prepared for the St. Louis CityPlan Commission, Mar. 1969); and in John F. Kain and John M. Quigley, "Evaluating theQuality of the Residential Environment," Environment and Planning 2 (Jan. 1970): 23—32.


TABLE 4-2Number of Unusable Home Interviews by Reason

ReasonSt. Louis

CitySt. LouisCounty

Nonresponse 108 7

Householder refused 127 19

Dwelling-unit vacant 191 1

Dwelling-unit demolished 14 0

Householder not English-speaking 5 0

Sample eliminated,inconsistent, incomplete, allother reasons 35 9

Number of households selected 1,583 206Number of acceptable samples 1,103 170

from the sample, leaving an analysis sample of 1,186 dwelling units andhouseholds.

The home-interview survey, which obtained data on both thehousehold and its dwelling unit, incorporated several novel features. Foreach dwelling unit, the interviewer was asked to ascertain the totalnumber of dwelling units in the structure, the number occupied, and thenumber of floors, and to rate the interior condition of common areas(halls, lobbies, and so on) and of the dwelling units. Each interviewer,moreover, was instructed to evaluate a number of specific features of thestructure/dwelling unit, i.e., ceilings, walls, floors, stairways, lighting,and windows. This procedure differs from that used in many othersurveys and in the 1940, 1950, and 1960 censuses of housing. Interview-ers in these surveys were instructed to provide an overall evaluation ofdwelling unit or structure condition. The 1970 Census of Housing, sinceit relied on self-enumeration, made no effort to obtain information ondwelling-unit condition. In the St. Louis Survey, interviewers wereasked to make detailed observations of specific features of each dwellingunit in the hope that they would be impressed with the importance of thequality evaluations and be discouraged from making hasty overall evalu-ations of the condition and quality of sampled dwelling units. In addition,they were asked to rank the quality of housekeeping on a scale of 1,excellent, to 5, very bad. This question was included to encourageinterviewers to make a distinction between the housekeeping and thephysical condition of sample dwelling units and structures.

In addition, for each sample unit, interviewers obtained the number


of rooms other than bathrooms, the number of bathrooms, and thenumber of sleeping rooms, and determined whether the unit had hotwater, central heating, and major appliances: air conditioner, refrigera-tor, stove, television set, telephone, washer, and dryer. The personsinterviewed were also asked whether they owned or rented their home,and owners were asked for the year in which they purchased theirpresent home, its purchase price, and their estimate of its current marketvalue. Questions were also included on the types and costs of homeimprovements made by owner-occupants in the previous eight years.Unfortunately, subsequent analyses of the home improvement dataindicated they were of very low quality. Regrettably, the survey did notobtain information on the method of financing owner-occupied homes oron financing terms.

Renters were asked both their current monthly rent and their rent inthe previous year. Moreover, interviewers determined whether heat,water, electricity, stove, refrigerator, and furnishings were provided bythe landlord and included in the rent. This information is necessary indistinguishing between structure rents and payments for utilities andfurnishings. Many analyses of the rental housing market are flawed bythe failure to distinguish between contract-rent and gross-rent payments.In this study, estimation of contract-rent corrections is an integral part ofsubsequent analysis. Renters were also asked to provide information onimprovements to their property, but preliminary analyses of these datasuggested they were even less reliable than the comparable informationfor owners.

In addition to descriptions of sample dwelling units, home inter-viewers obtained information on the size and composition of the family,and the age, sex, race, and labor-force attachment of all householdmembers. Interviewers were asked, wherever possible, to obtaindetailed breakdowns of income by source for each household member.These were aggregated to obtain the single measure of family incomeused in the analyses in subsequent chapters. The questionnaire alsoincluded questions on the household's perception of the neighborhoodand its problems, on household mobility, and on the health of familymembers. These data are not used in the analyses presented in thisbook. (The full home-interview questionnaire is reproduced as Appen-dix A.)

The Physical Blight Parcel Survey and the Environmental BlockFace Survey are the more novel parts of the survey design. These twosurveys, carried out by nine professional building inspectors working intwo-man teams, were designed to provide independent evaluations ofthe condition and quality of the exterior of the sample structure, of theparcel, and of adjacent structures and parcels. Because of their skill and


experience, and because each team could evaluate more structure exte-riors than a single home interviewer, it was hoped that the buildinginspectors would be able to obtain better and more consistent evalua-tions of the structures. The Physical Blight Parcel Survey also provideda useful consistency check on the evaluations of structure interiors bythe home interviewers, as well as an independent evaluation of thequality of conceptually distinct attributes of the bundles of housingservices.

The nine building inspectors were asked to provide quality ratingsfor specific attributes of the structure, as well as overall quality ratingsfor all sampled dwellings, in both the city and the county. Similarly, theywere asked to provide comparable quality ratings for adjacent parcelsand structures. Information was also obtained on the existence andquality of garages, driveways, fences, and other objective characteris-tics of the parcels. The questionnaire used in the Physical Blight ParcelSurvey is reproduced in Figure 4-5.

The second member of each building-inspector team performed theEnvironmental Block Face Survey. This survey, reproduced as Figure4-6, shared the parcel survey's concern with the quality and condition ofthe physical environment. However, its focus was much broader; itemphasized the local environment, particularly the block faces on bothsides of the Street containing the sample structure. The survey providedmeasures of the amount and type of nonresidential use; the generalcondition of the structures; the condition of streets, sidewalks, andstreet lighting; the amount of landscaping; and the amount of trash andlitter in the parcel area. This second team member also recorded anoverall quality-index rating for each block.

In addition to these three surveys, the analysis draws on a variety ofother published and unpublished data describing St. Louis neighbor-hoods. In particular, because of their oft-claimed importance in deter-mining the neighborhood quality and in affecting the residential choice ofhouseholds, a major effort was made to devise and obtain measures ofneighborhood services (both "objective" measures and those perceivedby residents). Special attention was given to neighborhood elementaryschools (both public and parochial) and to police protection.

Measurement of the quality of neighborhood services is a formida-ble undertaking. There are a number of difficult conceptual problemsassociated with such measurement and little clear-cut theory or system-atic empirical work to use as a guide. Where possible, both input andoutput data were gathered. However, the data collection and analysisemphasize output measures on the reasonable assumption that house-holds, in choosing a residential bundle, are not very interested in howmany policemen there are in the neighborhood but are very much


ST. LOUIS COMMUNITY RENEWAL PROGRAMPhysical Blight Parcel Survey

Residonliel San'.pI. Blocks

QUESTIONS AND CODFS

PARCEL IOEN1IFICRfl,014

I. Sampte number

LAND USE CLASSIFICATION

2. InnI. SIngle—detached5, Dayton3. 80cr4. flat5. Apartment6. Rocirnlng Hay.,7. Mihndconrmercia 1/

industrIal6. Orb.r

3, Ii eIsner ii 5, or .8, IrdIC,te tileniirtbertolnlc rr most closely apprnxl—mates roe mIxed land lunar

I. Rhitdehtta i/ted u.rrla I7. ReSt danulol/reraulS Rr8idrpniinl/resrasunant.

barI. flesta,rtuial/otIice,

ilnOtn, I toilet5. Retidentixi/.netat ServiceA. Re sidrlnlla I/unocCupied

flOe—re. 10.01101Rest dehuta i/tracking.

garafleI. ReOtdrnnal/narenauae

SlOt. cc

4. II the land rut, of the parcel I, tIringdent. 01 a dunn,.

rflrjltiflg frOm the non_rest deCal. I usa.net. items betore I rye.', Zanol.

A. Damns0. SmOkeC. UIOh flop.. level0. Euldeflile at mibratlon£ . Glare from lightingF. Ufldanlirabitiry at commercial

establi,

EXTERIOR EVALUAr1OS OF RESZIRNTIAL.SAMPLE STRUCTURE

S. Type Of COflItcuCtIncI. Masonry3.

3. Metal4, Coircneie1. Other

flat. tree. 6 througfl II accord nfl it thelOilOwinil SCalE

0. tern 00.5 nOn trill Ciii notObserve bit

I. Poor; drtOtiOr.I.d/ma in it fl4 Zr S e n eq lent rd

S. nit', maintenance. tipping3. Genteelly good Condition• . Excellent coil diLlon

S. Wait.A. SIraCtural 5flUflrZOeUU0. Paint or linhiti

7. POrch al

6. Outside Otairs/liepi9. WindowS

A. Window pane,0. Snh,r undo.C. Screens fsloen a,IOdOO,

it. Door.A. Entrance door (a)S. Screens/stUnt door Is)

It. Do,oflnpOot, ond gutters

It. Chin!,,

IS. Trio,

ii. Roof

IS. Intel root0. Nat observable

- Composition2. Asphalt shingLe3. Slur. nod/Or iii,4. Wood shingu.S. Oiflra

iS. P ate lilt oceroll eotrrior Condition ofthe residential acroiding Intie inIlo'aing scale;

Very poor; badlydttet iota r ed/u xc dv rd

2. Poor: qenerally bind,rv,hntnnonconagtnCtnd

I. Fair'. ammO. niainrenancesupping

.4. 0000: IOyitd. nnoioh,nane.odpqaoin

I. Encellenic000itiatn

EVALUA1'iON OP REMAiNDER OF PARCEL

LandA. Amorunr iii 5 once available tot

iar'4000pinqI. None2. Srratl amount3. Moderarn anouot4. LapIn OIrrOuOt

S. FtinOnint 03 landscaping0, SOt dppilcabieI. None Or miner2. M000rulr amount 01 grass.

alUobbery. or sitter reps3. Cons:denabieonacniof

C. Quality or lanrlscacingmaintenance

0. riot applicaNtI. Poor: d.terlOnai.ui/

ma mt to. ncr n FU 10Cr a0

2. Full; maintenanc.. Sipping3. Generally good canditioe4. E.rCelienl cocdiriaO

8. II accessory buildinfi Cal exist, ratetIle moot prOnintot.A. Cno.tructiOr

0. lIen do,. nOteolst or is notobsmtna bIt

I. Mactnt?1. Fran,S. Metti4. ConcreteI. Other

B, )j..0. Iron dona vol rout. otis

nor oS..et. SicI. fln.Idrfltlnl dcar'illflfl1. Garage3. Sinrona4. PtinyS. Abandoned6. Other

C. Condition0. 11Cm doeS nor exist, is

nor oboorriablaI . Poop: detect opaied/

maintecanncn pa lotted2. Ftitl Fnt,inietdfloe sIt pping3. GenerAlly good cOrndihion

4. Esceileni cOndition

IS. nences6. len a aes nut exist on is not

nba, rod bitI, Poor: dotenloratrd/

tsatripnaacr. negieorpdS. Fair; maIntenances tipping3, Generally good condition4. Excellent Connuulion

25. Watkaay.S. tram doe. not exist or

obsarna SIrI. Poor; d,iprloratnd/

naiflletianCe 1l.glrCIrdS. rail; inteinaocea IiPPioa3. Gerenalty good 000CItlOo4. Lnttltanttanrditton

St. Dniceony.0, Irert does not et;St is not

oboercobipP000 umauatacad or sighed

needs reptan,t,tenr5. Fair: hard •vrlocn, mend.

repair5, Good: nardaurfane. no dips

or holes4, Excellent 000diltan

22, Junk, trash, onrbageI. Poor: ObviOus

accurnolatinn Or Ilurar.tank and/or glaas

S. raft: moderair amount, oflitter VIsible, Ixut 00tins. or lank

3, Cuod: Very small xmoantU

4, CoonS ent:0c turret

23. Rate lint onerat: condturnn. on theretrta flexor at the pa rod amoral no I.the inltoalno .cUlt:

Very poor: badly detentoratudI. Poor: mountrirance negintielS. Faut: nra intenaoCea flppio,4. Good; adeqoatniy eta untainnd

0. Excellent condut tar

ADJACENT PROPcRr1ES

notion Inc lrnor Dl the simple Parcel,nb.rtne the parcel to your teN.

54. Doea tile parcel contain a PjceIyrpaidenriatnracrcrn P )i.yes; i.onl

CODING

interested in the number and seriousness of crimes. In short, we assumethat what concerns them is their personal safety and the security of theirpersonal property, not the problems of the police commissioner. Simi-larly, parents are presumably concerned with the quality of educationtheir children are likely to get, not how much it costs the city or how

1 2 3 4 5ABC DI F

6Al 7 8 9ABC

10 11AB

12 13 14 15 16 17 18ABC ABC

1920

FIGURE 4-5


S.pt.mb., i•17

25. lIthe parcel contain a partlyrt.,dtntIaI lInlelkIre, .01.41 IS Iretypo Or Itructurm?

• Slngl.-delaChcd1. 0091Cc3. ROts

4. rIdS.

S. Rooming House

2h. II he parcel ContaIn apurely ie.ld.nIlaIStrt,Ctare.

A, IS It pon I y rest dm1101?I.yrl; l.fluI

5. IndIcate It:. number Which moltclosely app'nolmatel theOr flOOrret 10*01161

I, Indu.trtll3. RetaIl. SCenIC.3. enterbllrtmert.bae4. 0111cc. pcole. .10061S. SchoolS. Church

Other EtaItlIC building OfSocial service

0. TruckIng, garage), Warebo.tse, lIthogeIt. Pack,cerreteryii. ParkIng lotIi. U00000000O na.—reSlden;ioIn. Vacart lotII Other

I. they. onolecce Cl ado.!.. lnhla1000.

lenin the 0.0—TeSidoOtial usage?Rat. It. m. StInt. Il.yra: 2.001

C. Odnil0. Snt,keC. High noiSe 1.0.1F. Evidence Of vibration

O - Glare Ton tIOhrinQU. Lack Of pfl000VI. 1161110 000ICSCI0O

27. Irate the 5oct42 lettering COtrdltltifl Utthe ItrucIsir. according io the 101101.1100

• Very pea.; badly d.terlO.altd/

7, POOr; 9.r.r.ally .0000,.uairt.flaflC. negleviod

i. rot,: 00.4.4. statOtecaticeSlIppling

4. Good: sound. 01 tolerance

adeQuateS. to OpUlent cotoilItTot

25. Retell.. ooerali condiTIon. nil III.rena mdv. 01 the parc,t occordlng tolb. 101100100 SCsi,;

• Very poor: badly deleylceoled2. Poor; maIntenance 0.91,0 led3. ralr: Tralniefla tCP Slipping

4. 00001: adequalely ndlnioin,dt. Excellent condItion

racing she (ton! TI the sample porch.Obatrue the gal theft Parcel to your 11201.

29. Do.. the parcel contolr 2 purelyresIdentIal structure? I l.A..', lanaI

30. II lit. porcel cenlalna purely

reSidentIal Structure.type ololtuclure?

• Slnqlrdelacired2. Duplex3, 5000

4. rlat5, &porim.ertt0, 500101100 Haul.

31. II the parcel p553 55.1 corialna purely

natgentlal Structure.

A. IS Ii paflly reSidenlIat?

2.0.5: 2.noI5. IndIcate lb. Tumble .ghIci, molt

clesely apPPOOimaie. the .550121Or LflSi!3C non-re,ldenIloI usUge.

ltdut IrIs I

2. RetolI, Irroice3. Rentouran I. emcee tamnment.

bar4. OffiCe. prnie..ioralS. SchoolA. Church7. 0th., public but Idling cc

Social serviceB, Truckittg. goroqe9. W.irhoasr, storage10. Perk, cemeteryII . Parking lot

Il. Onoocuttled nne!—reSl010IISIIi. VacantIi. Other

IS there eold000. ci later,. lrlluenr..reSultIng trom rIte

usage? gute lies.. helm. I .2.00)C. Odor.0. Smoket. HIOTI cot., levelr. Evidence Cl vIb,aIlO.0. GlOre Irate lightingU. Lack a0 patc000

I. Ironic coalettlat.

32. Rat. tile overall exterior enodilton 01itt eoituCture 0000idlitQ IT the(OllOwitig scale:

Very 000(1 badlY d.tell000letl/unsourOf

2. POt,; generolty .0Cfld.rn.inlenane. neglected

n. rant; e,alut.oanteslippIng

4. Gead: Sound. tltblrleu100eadequale

• ClnCtlleOiC 0.101:100

33. RAte lIlt Overall cccdltl&IS On the

remaInder 01 the parc,I .ccurdlmq I.lIne 101100100 ScOld

Yes-v poor; Sadly deirtlorsird1. POOr; toalctetoaflCe neglected3. role; ma lelenance 51190109

4. GOOd: ad,qaeL,iy tnatntolOeaS. Excellent 00,101000

34. Looking at th, entIre block tao., donsthe ofera II condItIon 0110, sampleporn,1 appear to bel

Con, lde'tObiy unt.. then tileocerane Ion lb. block lace

2. Generally the Salmon. toeaverage ICr Itte block lace

3. rae superior man It, akerogetOT th, block Tate

ALLEY

35. Type0. (40 olleyI. Straight0. Orlflierlenied

30. I. th, alley accessible toaurotnObllll 7lOerot • ppllcable; I—be.: 1.001

37. C001teootloo0. Not appIlcobieI. DIm

2. 0100.13. BrIck4. 11.12,00S. CITIcrete0. Other

38. COtiditlon0. CR1 applIcable

Pool; PariS, etori,draIned

a. raId ma lolenloceSlippIng

1. Good: hardSariaCT

30. O.c.r. I cl.artlmn.S.0. Nor applIcableI. Pot..: OboloaS

ncctumul.tloa CIlItter. 91405 400/0. lank

2. r.Ir: tlfodersleaiTIaart. Cl

litter, hIlt no cleat orluok

3, GOod: cr11 Small artoruot.on litter

4. Co COllect; 00 I

OFF STRUT PARKING

02. AdequacyI. NOrm2. Ira... goat, lIes. than I

trace par a_eli Inn untIl1. A 000uote Ii or mu.. .p.CeI

per 40,11mm loll)

2122232425 26 27 2829 30 31 32 333435363738 3940ABCDI FOHI ABCDE FGHI

DATE COMPI.ETED NAME OF INSPECTOR

many teachers in the school have a master's degree. Other people, suchas the elderly, may be concerned with taxes and have little interest in theschools.

Both output and input measures were obtained for each of thepublic and parochial elementary schools in the city of St. Louis. Atten-

ST. L

OU

ISC

OM

MU

NIT

Y R

EN

EW

AL

PR

OG

RA

ME

nvir

onm

enta

lB

lock

-F

ace

Surv

eyR

esid

entia

lS

ampl

e B

lock

s19

67

:.

Perc

ent o

f st

reet

fro

ntag

e in

:R

esid

entia

lC

omm

erci

alJn

dust

rlal

Publ

ic a

ndIn

stitu

tiona

lP

arks

and

Pla

ygro

unds

Vac

ant L

ots

Perc

ent o

f st

ruct

ures

obv

ious

ly v

acan

tC

ON

DIT

ION

OF

ST

RU

CT

UR

ES

Per

cent

of s

truc

ture

s w

hose

ove

rall

cond

ition

is:

Poor

; Det

erio

rate

d--O

bsol

ete—

—N

egle

cted

Fair

; Sou

nd--

Mai

nten

ance

Slip

ping

--R

epai

rs R

equi

red

Goo

d; S

ound

--W

ell M

aint

aine

dA

MO

UN

TO

F S

PA

CE

AV

AIL

AB

LE F

OR

LA

ND

SC

AP

ING

1.N

one

3.M

oder

ate

Am

ount

2.Sm

all A

mou

nt4.

Lar

ge A

mou

ntQ

UA

LIT

YO

F L

AN

DS

CA

PIN

G M

AIN

TE

NA

NC

E0.

Not

App

licab

le2.

Fair

; Mai

nten

ance

Slip

ping

Exc

elle

nt1.

Poor

; Litt

le M

aint

enan

ce3.

Goo

d; A

dequ

ate

Mai

nten

ance

DE

GR

EE

OF

TR

AS

H O

R J

UN

K A

CC

UM

ULA

TIO

N1.

Poor

; Obv

ious

acc

umul

atio

n of

litte

r, g

lass

and

junk

2.Fa

ir; S

mal

l am

ount

s of

litte

r vi

sibl

e, b

ut n

o gl

ass

or ju

nk3.

Goo

d; N

o ob

viou

s lit

ter,

gla

ss o

r ju

nk in

str

eets

or

yard

sC

ON

DIT

ION

OF

SID

EW

AL

KS

0.N

o Si

dew

alks

2.Fa

ir; N

eed

Min

or R

epai

rs1.

Poor

; Nee

d of

Rep

lace

men

t3.

Goo

d; I

n G

ood

Con

ditio

nA

DV

ER

SEIN

FLU

EN

CE

S R

ELA

TIV

E T

O R

ES

IDE

NT

IAL

US

E

L. 1

OD

DB

LOC

K

—k-

—A

-

AD

DR

ES

S R

AN

GE

EV

EN

BLO

CK

NU

MB

ER

LAN

D U

SE

ST

RE

ET

NA

ME

OD

D B

LOC

K N

UM

BE

R L

C m 0)

EV

EN

BLO

CK

—

LI

LILI

LII

LI

LILI

Obj

ectio

nabl

eH

igh

Noi

seO

bjec

tiona

ble

Evi

denc

e of

Gla

re F

rom

Chi

ldre

n Pl

ayin

g in

Odo

rsL

evel

Smok

eV

ibra

tion

Lig

htin

gor

Aro

und

Stre

et

1Y

es2

No

L! Y

es2

No

IY

es2

No

ri Y

es2

Nu

J2

Na

J[1

Yes

2 N

o j

ES

TH

ET

IC C

ON

SID

ER

AT

ION

SD

oei

ther

or

both

blo

ck f

aces

con

tain

any

unu

sual

or

uniq

ue e

sthe

tic f

eatu

res

whi

ch e

nhan

ce th

ech

arac

ter

of th

e ar

ea?

Arc

hite

ctur

eL

ands

capi

ngO

ther

IY

es2

No

J1

Yes

2N

o1

Yes

2N

o

CO

ND

ITIO

N O

F S

TR

EE

T1.

Poor

; Itn

surf

aced

or

surf

ace

need

s re

plac

emen

t2.

Fair

; Har

d su

rfac

e——

need

s m

inor

rep

air

3.G

ood;

Har

d su

rfac

e——

no h

oles

dips

or

Irre

gula

ritie

sC

ON

DIT

ION

OF

CU

RB

S A

ND

GJT

TE

RS

0.N

ocu

rb a

nd g

utte

r1.

Poor

; Str

uctu

ral c

ondi

tion

poor

——

maj

or r

epai

rs o

r re

plac

emen

t req

uire

d2.

Fair

; In

need

of

min

or r

epai

rs3.

Goo

d; S

truc

tura

l con

ditio

n go

od—

—no

hol

es, c

rack

s, e

tc.

AD

EQ

UA

CY

OF

ST

RE

ET

LIG

HT

ING

1.N

o st

reet

ligh

ts2.

Exi

stin

g st

reet

ligh

ts a

re n

ot a

dequ

ate

3.E

xist

ing

stre

et li

ghts

are

ade

quat

eT

RA

FF

IC C

ON

DIT

ION

SP

erce

ntof

cur

b sp

ace

norm

ally

ava

ilabl

e fo

r pa

rkin

gPa

rkin

g Pr

ohib

ited

Park

ing

Park

ing

Tra

ffic

Lig

htSt

op S

ign

atB

us S

top

onon

Bot

h Si

des

Proh

ibite

d O

nePr

ohib

ited

Peak

at E

ither

Eith

er e

nd o

fE

ither

All

Peri

ods

Side

All

Peri

ods

Tra

ffic

Per

iod

End

of

Blo

ckB

lock

Blo

ck F

ace

2Nol

lives

2Nol

lives

lives

2Nol

[IVes

Doe

sIt

app

ear

that

the

amou

nt o

f co

mm

erci

al tr

affi

c on

this

str

eet i

s:1.

Hea

vy2.

Mod

erat

e3.

Litt

le o

r no

neO

VE

RA

LL

CO

MB

INE

D B

LOC

K F

AC

E R

AT

ING

1.B

ad; M

ajor

deg

ree

of s

truc

tura

l def

icie

ncie

s w

hich

are

bey

ond

repa

ir--

maj

or d

efic

ienc

ies

in m

aint

enan

ce, l

ands

capi

ng, a

ccum

ulat

ion

of tr

ash

or a

dver

se in

flue

nces

.2.

Poor

; Som

e st

ruct

ural

def

icie

ncie

s——

maj

or d

efic

ienc

ies

in m

aint

enan

ce, l

ands

capi

ng,

accu

mul

atio

n of

tras

h or

adv

erse

infl

uenc

es.

3.F

air;

Mod

erat

e de

fici

encl

esin

mai

nten

ance

, lan

dsca

ping

, acc

umul

atio

n of

tras

h or

adve

rse

infl

uenc

es.

4.G

ood;

Onl

y m

inor

def

icie

ncie

s.5.

Exc

elle

nt:

No

defi

cien

cies

——

evid

ence

of

cons

ider

able

spe

ndin

g.

0D

AT

EC

OM

PLE

TE

D__

____

____

____

____

__N

AM

E O

F IN

SP

EC

TO

R__

____

____

____

____

____

____

____

____

____

_


dance districts of each school were determined, and these measureswere coded to each dwelling unit in the city sample. Output was mea-sured by three achievement-test scores: reading, literary-writing, andmath. Identical tests were used for both types of schools, but in theparochial schools they were given to the seventh grade, whereas public-school scores are for eighth-grade students. The input measures werestudent-teacher ratios and student-classroom ratios. In addition, thepercentage of students and teachers that were nonwhite, the school'sage, and a measure of the condition of the physical plant were obtainedfor public elementary schools.

It was not possible to obtain at reasonable cost input measures ofpolice protection for each neighborhood. Thus, only output measures—the number of major crimes and the number of minor crimes per Paulyblock6—were collected. These were also coded to each sampled dwell-ing unit.

Summary statistics of these indexes of school quality and policeprotection are presented in Table 4-3. From the data shown it is clearthat there are important differences in public and parochial schools, andin the levels of criminal activity. Both factors are widely believed toexert a strong influence on a household's choice of residential location.

Obviously, it would be desirable to have comparable data for thecounty samples. Many explanations of the decline of central cities havefocused on the better schools and lower level of criminal activity foundin suburban areas. Unfortunately, comparable data enabling us to evalu-ate these claims were not available at reasonable cost for those unitslocated outside the central city. Moreover, even if the figures wereavailable, it is doubiful that the impact of these two factors and severalother county—central-city hypotheses could be adequately tested withthe data at hand, given the small size of the county subsample. Thesubsample of St. Louis county households is invaluable as an extensionof the city sample, but it is too small and incomplete to permit testingsome of the most important hypotheses about central-city—suburbancompetition for residents.

Several kinds of data were obtained from the city and countyassessors' records for each of the sample dwelling units. The mostvaluable of these data were the assessed value of land and structures,structure age, parcel area, building area, and type of construction. Thesedata were supplemented by professional real-estate appraisals obtainedfor a subsample of five-hundred parcels within the city. These apprais-

6Pauly blocks are geographic areas of roughly equal sizes named after LieutenantPauly of the St. Louis Police Department, the developer of the statistical reporting systemused in recording and tabulating crime data in St. Louis.

TA

BLE

4-3

Cha

ract

eris

tics

of N

eigh

borh

ood

Sch

ool,

Crim

e, a

nd A

cces

sibi

lity

Mea

sure

s

Var

iabl

eD

escr

iptio

nM

ean

Val

ueSt

anda

rdD

evia

tion

Min

imum

Max

imum

Publ

ic s

ervi

ces

(city

onl

y)M

ajor

cri

mes

By

Paul

y bl

ock

in 1

967

97.9

981

.45

2.00

432.

00O

ther

cri

mes

62.8

248

.60

0.00

349.

00Pu

blic

Sch

ools

8th

grad

eR

eadi

ng s

core

7.83

1.21

6.70

9.20

Lan

guag

e sc

ore

8.49

1.27

6.80

9.90

Mat

hem

atic

s sc

ore

7.83

1.22

6.70

9.90

Paro

chia

l Sch

ools

7th

grad

eR

eadi

ng s

core

7.83

0.62

6.70

9.00

Lan

guag

e sc

ore

8.53

0.64

7.10

9.80

Mat

hem

atic

s sc

ore

7.75

0.75

6.50

9.60

Dis

tanc

e an

d ac

cess

ibili

tym

easu

res

(ent

ire

sam

ple)

Dis

tanc

e fr

om C

BD

Mile

s5.

155.

230.

0512

.10

Acc

essi

bilit

y to

em

ploy

men

tIn

dex

valu

e7.

882.

300.

8112

.53

Acc

essi

bilit

y to

per

sona

lbu

sine

ssIn

dex

valu

e2.

400.

750.

284.

05A

cces

sibi

lity

to s

hopp

ing

Inde

x va

lue

1.89

0.53

0.33

4.11

Acc

essi

bilit

y to

soc

ial c

onta

ctIn

dex

valu

e1.

450.

320.

282.

05A

cces

sibi

lity

to r

ecre

atio

nIn

dex

valu

e1.

360.

420.

322.

35A

cces

sibi

lity

to n

on-h

ome-

base

d tr

ips

Inde

x va

lue

2.37

0.51

0.44

3.78

0 (.0


als, which included estimates of the current market value of land andimprovements, and quality ratings of structure condition, provided anextremely valuable check on the accuracy and consistency of both theassessments and the owners' estimates. The estimate of market value forsingle-family units used in subsequent analyses is a composite of thesethree sources. The procedures used are described in Appendix B. Thequality ratings provided by the appraisers provided still another indepen-dent evaluation of the quality of sampled structures.

Six measures of auto accessibility were available for each observa-tion. These are accessibility of each dwelling unit to: (1) employment, (2)personal business, (3) recreation, (4) shopping, (5) social contact, (6)non-homebased trip destinations. In addition, each sampled dwellingunit was located on a grid (X, 1') coordinate system. This permitted eachunit to be located in terms of its distance from any location within the St.Louis metropolitan area. Extensive tests of these accessibility measureswere performed. All are highly correlated. Distance from the centralbusiness district, the simplest of these measures of accessibility, is usedin most of the analyses reported here.

THE SAMPLE DESIGN

The samples used in our analyses—approximately 1,500 privatelyowned dwellings in St. Louis city, and approximately 200 in St. LouisCounty—are based on a relatively complicated sample design. The citysubsample was drawn using a stratified cluster sample, in which theprobability of selection was higher for households in low-income neigh-borhoods than for those in high-income neighborhoods. The suburbansubsample is a stratified random sample of St. Louis County dwellingunits, which employs different sampling rates for single and multifamilyhouseholds. In addition, 50 public-housing units were sampled. How-ever, these units are not included in our analyses.

THE CITY SUBSAMPLE

Sample dwelling units for the city subsample were drawn from oneof four sampling strata defined in terms of 1960 census-tract medianincomes. They are:

Stratum I—Median family income of $3,000 or less;Stratum Il—Median family income of $3,00l—$5,000;Stratum Ill—Median family income of $5,001—$6,000;Stratum IV—Median family income of more than $6,000.


A cluster sampling technique, which insured that interviewerswould have several addresses located near one another, reduced inter-viewing costs. The desired geographic clustering was obtained by use ofa two-level method of sampling: (1) a predetermined number of blockswas randomly selected within each strata, and (2) a predeterminednumber of dwelling units was randomly selected within the sampleblocks of each strata.

The desired number of sample blocks in each stratum was com-puted by dividing the number of sample dwelling units to come from astratum by the desired average per block for that stratum. A compromisebetween coverage and economy in data collection dictated an overallaverage of 3 samples per block. However, because of the heavy sam-pling in Stratum I and Stratum U, the average number of sample dwellingunits per subblock for these strata was increased to 4.3 and 3.5, respec-tively. Since several city blocks in each stratum contained no residentialland use, it was necessary to oversample other blocks to compensate fortheir omission.

From City Plan Commission records, each city block was allocatedto the appropriate income stratum, and city blocks within each stratumwere sampled separately by a random-interval sampling procedure.Table 4-4 shows the results of this technique.

The second phase of selecting the private dwelling-unit sample—theselection of approximately fifteen-hundred residential dwelling unitsfrom within the sample blocks—was conducted in a similar manner. Thenumber of dwelling units on each sampled block was obtained from CityPlan Commission records, and the sampling rates for each stratum weredetermined by dividing the total number of dwelling units within thesample blocks of each stratum by the desired number of dwelling-unitsamples.

The sample dwelling units were then selected within each incomestratum by applying the sampling rate to all dwelling units, beginningwith the northeast corner of the first sample block (the one with thelowest City Plan Commission identification number). Sanborn mapswere used to obtain the correct street addresses and apartment numbersof the samples. When the last unit had been selected in the first block,the sampling rate was continued onto the next sample block in numericorder until all sample blocks in each stratum had been covered. Thismethod assured that all dwelling units within the sample blocks had anequal probability of selection. Table 4-5 shows the results of the sam-pling procedure.

It is apparent from Table 4-5 that the average sampling rate for theentire city is one sample for each 159 private dwelling units. By strata,however, the rates range from a high of one sample for each 66 private

TA

BLE

4-4

Dis

trib

utio

n of

Priv

ate-

Hou

sing

Sam

ple

Blo

cks

Des

ired

Des

ired

Min

imum

Act

ual N

umbe

rT

otal

Num

ber

Num

ber

ofA

vera

geD

esi

red

Num

ber

of S

ampl

eof

City

Dw

ellin

g-pe

rof

Sam

ple

Blo

cks

Blo

cks

Uni

t Sam

ples

Blo

ckB

lock

sSe

lect

ed

Stra

tum

I65

715

24.

038

41St

ratu

m I

I1,

520

563

3.5

161

168

Stra

tum

III

1,68

842

93.

013

814

9St

ratu

m I

V2,

061

356

3.0

123

132

Tot

al5,

926

1,50

0—

460

490

I')

TA

BLE

4-5

Dis

trib

utio

n of

Priv

ate-

Hou

sing

Sam

ple

Dw

ellin

g U

nits

Num

ber

ofD

esir

edA

ctua

l Num

ber

Ave

rage

Num

ber

Tot

al N

umbe

rSa

mpl

eN

umbe

r of

of S

ampl

eof

Sam

ple

of P

riva

teB

lock

sSa

mpl

eD

wel

ling

Uni

tsD

wel

ling

Uni

tsD

wel

ling

Uni

tsSe

lect

edD

wel

lingU

nits

Sele

cted

per

Blo

ckSa

mpl

ing

Rat

e'

Stra

tum

I11

,124

4115

116

94.

166

Stra

tum

II

72,8

4416

856

355

93.

313

0St

ratu

m I

II76

,911

149

429

432

2.9

178

Stra

tum

IV

81,8

4813

235

636

62.

822

4T

otal

242,

727

490

1,50

01,

526

3.1

159

"Tot

al n

umbe

r of

pri

vate

dw

ellin

g un

its"

divi

ded

by "

num

ber

of s

ampl

e dw

ellin

g un

its."


dwelling units in the lowest-income census tracts to a low of one sampleper 224 private dwelling units in the highest-income tracts.

SAMPLING PROCEDURE FOR ST. LOUIS COUNTY

The procedure used to select the 206 sample dwelling units fromwithin St. Louis County was somewhat different from that used for thecity. The suburban subsample was designed to supplement the citysample in Income Stratum IV and to provide a better description of thefull range of residential alternatives available in the urban area.

The first phase of the county sampling procedure involved theselection of a sample of parcels containing residential dwelling unitsfrom an alphabetical list of county blocks and their associated parcels inthe County Assessor's Office. A random-interval sampling procedurewas used to select 287 sample parcels containing residential dwellingunits. Then, the number of dwelling units associated with each of thesesample parcels was determined from county assessment records.

The. second phase of the sampling procedure involved the selectionof dwelling units from two subsamples within the 287 sample parcels: asubsample of 184 dwelling units from the 256 one- or two-family-struc-ture parcels and a subsample of 22 dwelling units from 31 multiple-unit ormixed-land-use parcels. Table 4-6 summarizes the selection results.

A BRIEF NOTE ON WEIGHTING

It is obvious from the preceding discussion of the sampling proce-dures used to collect the home-interview data analyzed in the course ofthis book that the sample observations represent different numbers ofhouseholds. This raises the question of how the sample should be treatedin subsequent analyses. Specifically, the issue is whether, in the statisti-cal analyses that follow, the sample observations should be weighted bythe sample weights shown in Tables 4-5 and 4-6 or whether they shouldbe weighted uniformly. As the statistics in Table 4-7 indicate, the choicemakes a considerable difference in the computation of sample means.For both the renter and owner subsamples, the unweighted means havea larger proportion of black households, lower incomes, and lower housevalues and rents than the corresponding population estimates.

If the objective of the analyses that follow were to estimate popula-tion parameters for the city and county of St. Louis, the choice would bestraightforward, and individual observations would be weighted. Thecase is less clear, however, in the statistical analyses presented in

TA

BLE

4-6

Sum

mar

y of

the

Cou

nty

Res

iden

tial-D

wel

ling-

Uni

t Sam

ple

Est

imat

edN

umbe

rD

esir

edof

Num

ber

ofN

umbe

r of

Num

ber

ofPr

ivat

eSa

mpl

eSa

mpl

eSa

mpl

eSa

mpl

ing

Dw

ellin

g U

nits

Parc

els

Dw

ellin

g U

nits

Dw

ellin

g U

nits

Rat

e'

From

1-

and

2-fa

mily

dw

ellin

gun

its19

9,62

625

618

018

41,

085

From

mul

tiple

and

mix

eddw

ellin

g un

its57

,549

3120

222,

616

Tot

al25

7,17

528

720

020

61,

248

'"E

stim

ated

num

bero

f pr

ivat

e dw

ellin

g un

its"

divi

ded

by "

num

ber

of s

ampl

e dw

ellin

g un

its."

01


TABLE 4-7Unweighted and Weighted Means for Renters and Owners in St. Louis Cityand County

Renters Owners

Unweighted Weighted Unweighted Weighted

Proportion blackIncomeRentValue

.45

$5,395$63.31

—

.24$6,753$80.75

—

.18$8,618

—

$16,512

.07$10,218

—

$18,973

subsequent chapters. Although a good argument can be advanced forusing the sample weights in these regressions, we have generally fol-lowed the opposite convention. In Appendixes D and E simple expendi-ture models are presented using both weighted and unweighted regres-sions. With these two exceptions, the regressions we present are basedon unweighted observations.

If the individual equations estimated were correct specifications oftrue structural equations, equal weighting of observations would haveonly small effects on the estimation of individual parameters. The ordi-nary least-squares (unweighted) procedure would be less efficient, butthe estimated coefficients would be unbiased.

SUMMARY

This chapter provides some pertinent background information onSt. Louis, Missouri, and its environs. It also includes: a brief descriptionof the survey instruments used to gather information for a sample of1,583 dwelling units and households in the city of St. Louis and for aseparate sample of 206 dwelling units and households in St. LouisCounty; and a discussion of collateral data obtained from assessmentrecords for St. Louis city and County, from public and parochial schooldistricts, from the St. Louis Police Department, and from the East-WestGateway Council, as well as from the U.S. Census. (The survey instru-ments used in the study are included as Appendix A.)

The sample of dwelling units located in the central city was chosenby a stratified, random-sampling technique in which the probability ofselection for any dwelling unit varied inversely with the (1960) medianincome of the census tract. The smaller suburban sample was chosen bya random selection of blocks in St. Louis County and by a random


selection of dwelling units (stratified by single detached and multifamilyunits) within each block. The selection and survey process resulted in asample of 1,186 dwelling units and households. With the exception of theanalysis in Chapter 6, the analyses contained in the rest of this volumeare based upon this sample of households and dwelling units.

The surveys used in this study had several novel features. Perhapsthe most significant of these was the use of overlapping surveys carriedout by home interviewers and building inspectors. These surveys weredesigned to elicit information about the dwelling unit, the neighborhoodenvironment, and the household. The objective was to obtain detailedand comprehensive information about the broad range of living condi-tions (or the housing services, broadly defined) associated with eachsample dwelling unit. In particular, a serious effort was made to describeand to quantify the qualitative as well as the quantitative attributes ofdwelling units, structures, parcels, and their associated block faces andneighborhoods.

Besides its small size, the most serious weakness of the sample usedmost of the analysis presented in subsequent chapters (we suspect

that as a general rule all samples are considered too small by theresearchers who use them) is the lack of an adequate sample of house-holds and dwelling units from the suburb orfrom communities surround-ing the central city. Because of the small suburban subsample andbecause some information was simply unavailable for any suburbanhousing units, at several points in the subsequent analysis we are unableto address adequately a number of crucial issues. On the other hand, thesample available for empirical analysis is extremely rich in its descrip-tion of households and dwelling units, and its heavy oversampling ofpoverty households permits what is probably the most comprehensivequantitative analysis of the effects of social discrimination on the hous-ing-market behavior of black households.

characteristics of the study area and of the data · 2020. 3. 20. · table 4-1 characteristics of...

Documents