USF Libraries
USF Digital Collections

Transit accessibility and labor force participation rate of at-risk groups

MISSING IMAGE

Material Information

Title:
Transit accessibility and labor force participation rate of at-risk groups Dade County
Physical Description:
1 online resource (iv, 18, 6 leaves). : ;
Language:
English
Creator:
Thompson, Gregory Lee, 1946-
United States -- Dept. of Transportation. -- University Research Program
Florida State University -- Dept. of Urban and Regional Planning
National Urban Transit Institute (U.S.)
Publisher:
University of South Florida, Center for Urban Transportation Research
Available through the National Technical Information Service
Place of Publication:
Tampa, Fla
Springfield, VA
Publication Date:

Subjects

Subjects / Keywords:
Local transit -- Florida -- Miami-Dade County   ( lcsh )
Commuting -- Florida -- Miami-Dade County   ( lcsh )
African Americans -- Transportation -- Florida -- Miami-Dade County   ( lcsh )
Hispanic Americans -- Transportation -- Florida -- Miami-Dade County   ( lcsh )
Choice of transportation   ( lcsh )
Genre:
bibliography   ( marcgt )
technical report   ( marcgt )
non-fiction   ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 16-17).
Funding:
Prepared by Florida State University, Dept. of Urban and Regional Planning and sponsored by the University Research Institute Program, U.S. Dept. of Transportation under contract no.
Statement of Responsibility:
Gregory L. Thompson.
General Note:
"June 1997."
General Note:
Final report.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 029179060
oclc - 754659432
usfldc doi - C01-00136
usfldc handle - c1.136
System ID:
SFS0032246:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 2200421Ia 4500
controlfield tag 001 029179060
005 20110926162022.0
006 m o d
007 cr bn|||||||||
008 110926s1997 flu obt 000 0 eng d
datafield ind1 8 ind2 024
subfield code a C01-00136
035
(OCoLC)754659432
040
FHM
c FHM
043
n-us-fl
090
HE5634.M54 (ONLINE)
1 100
Thompson, Gregory Lee,
d 1946-
0 245
Transit accessibility and labor force participation rate of at-risk groups
h [electronic resource] :
b Dade County /
Gregory L. Thompson.
260
[Tampa, Fla. :
University of South Florida, Center for Urban Transportation Research ;
Springfield, VA :
Available through the National Technical Information Service,
1997]
300
1 online resource (iv, 18, [6] leaves).
513
Final report.
Oct. 1995-June 1997.
536
Prepared by Florida State University, Dept. of Urban and Regional Planning and sponsored by the University Research Institute Program, U.S. Dept. of Transportation under contract no.
DTRS 93-G-0019.
500
"June 1997."
5 FTS
504
Includes bibliographical references (leaves 16-17).
588
Description based on print version record.
650
Local transit
z Florida
Miami-Dade County.
Commuting
Florida
Miami-Dade County.
African Americans
x Transportation
Florida
Miami-Dade County.
Hispanic Americans
Transportation
Florida
Miami-Dade County.
Choice of transportation.
710
United States.
Dept. of Transportation.
University Research Program.
2
Florida State University.
Dept. of Urban and Regional Planning.
National Urban Transit Institute (U.S.)
776
i Print version:
Thompson, Gregory Lee, 1946-
t Transit accessibility and labor force participation rate of at-risk groups.
[Tampa, Fla. : University of South Florida, Center for Urban Transportation Research ; Springfield, VA : Available through the National Technical Information Service, 1997]
w (OCoLC)37771666
773
Center for Urban Transportation Research Publications [USF].
4 856
u http://digital.lib.usf.edu/?c1.136
FTS
y USF ONLINE ACCESS
951
10
SFU01:001967960;
FTS



PAGE 1

Transit Accessibility and Labor Force Participation Rate of At-Risk Groups: Dade County Final Repon Principal Investigator Gregory l. Thompson June 1997 F l orida State Un i versity Department of Urban and Regional Planning Tallahassee, Florida 32306 -2030 (904) 644-4510 office (9041 644-6041 fax

PAGE 2

DiSCLAiMI:R The contents of this report reflect the views of the authors who are responsible for the facts and the accuracy of the information presented herein. This document is disseminated under the sponsorship of the De.partment of Transportation, Un.iverslty Research Institut e Program in the in terest of information exchange. The U. S. Government assumes no liability for the contents or use thereof.

PAGE 3

'RCHNICAL REPORT STANDARD mt PAGE .. ....,, '"' NUTI3FSU 2 AeC$UIOil t(o, 3 Nu. 4 T ille SIIIXIIIC 6 Rcpolt June 1997 i Transit Accessib ility and Labor Force Participat i on Rate of At Risk Groups: Dade Coun t y I 6. Performr.g 1 Aul.hOrt Gregory L. T h ompson 8 P$tro:MII'Ig OIO&IliltiCXII'I No, t, 0 fOllnl.r.lltlo n Mme Mtl Alll1flml 10, VJO11JJt'.nt acces slb ililY, labor force participation rate, spatial mismatc h theory, transit Ava ilable to the pub li c thro ug h the National Technical Information mode split. racial/ethnic S er v i ce INTIS!. 5285 Port Royal Road, Spr i ngfie l d, VA 22181, ph 1703) 487-4650 1 9 C l ou:.;l lolll'ils <$0111 20. Securit y lOt 21 22. P..IC$ f iorl r I" 1/UU. 1

PAGE 4

Table of Contents Table of Contents 0 P r eface . . . . . . . . . . . . . . . . . . . . . . . . . ii Acknowledgments . . . . . . . . . . . . . . . . . i i i ABSTRACT . . . . . . . . . . . . . . . . . . . . . . i v INT RODUCTION ............. . . . . . . . . . . . . . . 1 LITERATURE REVIEW . . . . . . . . . . . . . . . . . 1 Early Suburban Transit Demonstrations . . . . . . . . . 2 Analysis o f Watts Riots . . . . . . . . . . . . . . . 3 Kerner Commission . . . . . . . . . . . . . . 4 Spatial Mismatch Theory . . . . . . . . . . . . . . . . . . . 5 DADE COUNTY STUDY OF IMPORTANCE OF TRANSIT ACCESSIBILITY . . . . . 7 Hypothesis and Means for Testing . . . . . . . . . . . . . . 7 Study Site . . . . . . . . . . . . . . . . . . . . . . 7 Study Design . . . . . . . . . . . 8 Sample Size . . . . . . . . . . . . . . . . . . 8 Accessibility Ind ices . . . . . . . . . . . . . . . . . 8 Other Var i ables and Weightin g . . . . . . . . . 10 Results . . . . . . . . . . . . . . . . . . . . . . 1 1 DISSCUS SION AND CONC L US IONS . . . . . . . . . . . . . . . 14 REFERENCES ............. ..... ... .... . . . . . . . . 16 TABLES . ............................... ......... .... 18

PAGE 5

Preface This project proceeds from resulis of an earlier National Urban Transit Institute study that used Sacramento data to determine transit trip production and attraction potenti al in suburban census tracts. That study showed that potential tr. ansit ridership was independent of the quality of transit service offered and depended upon demographic and land use characteristics of the census tracts. I n contrast, observed transit ridership to and from each tract materialized from potential in direct proportion to transit accessibility of the tract to destinations to which potential transit users wished to go. In year one we conducted another study based on Orlando data that aggregated assessor parcel files to the census tract level and merged them with Census STF3 f i les. T his enabled us to use land use variables in mod els of travel behavior, and we conducted trial esti mations of simple models. In this study we examine the relationship of transit accessibility ard employment in Dade County, Florida. II

PAGE 6

Acknowledgments This project is made possib le through a grant from the U.S. Department of Transportation, University Research I nst i tute Program. Their support is gratefully acknowledged. I also would like to thank the Flor ida Department of Transportation Public Transit Office and Systems Planning oifice for assistance, and the Metro Dade Transit Authority and the Metro Dade Metropoli tan Planning Organization for providing data, computer models, and assistance, without which this project would not have been possible. Many individuals a lso provided invaluab le help. I would like to thank Mr. Michael Moore of the Metro Dade Metropolitan Planning Organization for providing files for the 1990 Dade County transportation model, Mr. Wilson Fernandez of the Metro Dade Transit Authority and Mr. Harry Gramling of Flor ida Department of Transpo rtation's Systems Planning Office for helping us to debug and run the model, Mr. Whitt Blanton of JH K & Associates in Orlando for providing us the 1990 Census Transportation Planning Package for Dade County, student i ntern and Transit Fellow Mr. Kevin Ti lbury for running the Dade County transportation model, and f ina lly student i nterns Messrs. Jon Sewell, Melvin Mitchell, and Matt Click for researching spatia l mismatch theory. I also would like to acknowledge the important ro le of a new program, the Transit Fellow Program, in conducting this work. The Florida Department of Transportation's P ublic Transit Office and the Department of Urban and Regional Planning at Florida State University set up a program in 1995 to entice bright students into the f ield of public transportation with tuition fee waivers and stipends for living support. Students named Transit Fellows are required to emphasize transit planning in their studies for a Master s of Science Deg ree in Urban and Regional Planning, and they are required to intern with transit authorities in Florida. The Transit Fellow Program placed one of its f irst fellows, Kevin Tilbury, with the Metro Dade Trans i t Authority's management in tern program during the Summer 1996. During the following two semesters, Mr. Tilbury was able to use contacts deve loped during his internship to obtain the Metro Dade transportation simulation for 1990 an d r un it through to completion thereby providing transit travel times for the accessibility indices central to this study. iii

PAGE 7

Transit Accessibility and Labor Force Partic ipation Rate of At-Risk Groups: Dade County ABSTRACT This report examines whether transit accessibility to jobs from indiv i dua l traffic analysis zones has an impact on employment of different racial/ethnic groups. Demonstrations of cross -town bus routes making suburban employment more accessib l e have been well-used by people traveling to suburban jobs, supporting the i dea that transit accessibility may make a difference. Earlier literature on spatial mism a tch theory suggests that t ransi t accessibility should make a difference in unemployment rates for Afri can-Americans confined to inner c ity ghettos. In contrast, more recen t literature uses theory to discount the importance of accessibility in l owering unemployment, and empirical observation supports the theory Because in all of these studies transit accessibility was not measures precisely, this study attempts to do so. Accessibility is measured as employment in all regional traffic analysis zones, each inve rsely weighted by the square of the door-to-door transit travel between the zone where the employment is and the zone whose accessibility is being measured. This measure of accessibility is then used as one of several exp lanatory variables in models of Afri can-American. Hispanic white, and non Hispanic white unemployment, work trip transit mode split, and automobile ownership in traffic analysis zones for Dade County Florida. This research finds that transit accessibility does not explain Afr ican-American unemp loymen t directly but on the margin it explains African American transit mode split to work, and it has a very import ant inverse relationship in models of automobile ownership in Dade County i v

PAGE 8

Transit Accessibility and Labor Force Participation Rate of At-Risk Groups: Dade County INTRODUCTION The McCone Commission {Governor's Commission, 19651 investigating the causes of the 1965 Watts riots identified insufficient public transpo rtation to suburban employment sites as a contributing cause for the high unemployment rate of Americans. The McCone findings inspired policy initiat i ves to improve transit services for what became known as the transit disadvantaged. Federal policy also reflected this concern by requiring transit operators receiving federal capital and operating grants to show the extent to which the residences of targeted groups lived close to transit routes. Results of such initiatives proved disappointing. While improved transit service to suburban employment sites did increase African American employment,. the newly employed used their new income to purchase autos and desert transit. In some areas it suspected that nearby transit routes were irrelevant to many transit-dependent groups, who wanted to trave l to places other than downtown. In many U.S. c i ties the range offered by CBO-oriented systems and new investment projects was so limited that public transportation was l argely as irrelevant to transit disadvantaged groups as to other sectors of society {Warner, 1972, pp. 142-144). Despite recognition that the range of destinations offered by transit from different neighborhoods is important to the lives of disadvantaged groups, there has been no formal study of this topic until recently In recent years the transit disadvantaged have been redefined as the mobility-impaired. Federal policy has focused on how accessib le transit services are to the mobility-impaired using the limited definition of accessibility as access to bus and train stops. It ignored the equally important aspect of accessibility, which is whether transit service will take passengers to where they want to go once they are on board buses or trains. In recent work {Thompson 1997a and 1997b) I have begun to study how important a range of destinations really is to transit users. This work concludes that the range of destinations available to potential users influences transit usage. It also shows that range of destinations that users can reach is influenced both by transit route configuration and land use con f igurat i on. Building on my recent past and on-going work, this project goes a step further. It examines the extent to which a measure of transit accessibility, based on how easily transit users from any g i ven area can reach jobs in the entire region, affects the unemployment rate of African Americans, Hispanic whites, and non-Hispanic whites in a major metropolitan region, Dade County, Florida. The hypothesis tested is that areas of urban regions with more transit accessibility experience h igh er labor force participation rates among poore r workers. In order to gain greater insight into possible links between transit accessibi lity and employment of different racial/ethnic grups this study also tests two other hypoth eses. One is that transit accessibility is important to t r ansit ridership of different groups from different parts of the region. The other is that transit accessibility i s inversely related to auto ownership of different rac ia l/ethnic groups. LITERATURE REVIEW

PAGE 9

The idea that improved public transportation between central city ghettos and suburban employment centers would improve employment prospects for minority groups stems from two r elated sources. One source was an analysis of the Watts, California, riot of summer 1965, w hich identified inadequate public transportation from the ghetto to suburban jobs as a contributing factor. Following the Watts riots, other urban rac ia l riots occurred, and from further study of their causes a mor e formal theory was put forward to explain the role of transportation in racial grievances. This was the spatial mismatch theory, which further strengthened the i dea that improved transportat ion to suburban employment centers would improve the social and economic well-being of African American and other minority groups. The following sections summarize the early demonstrations, transportation concl usions drawn from the racial riots of the l ater 1960s, the spatial mismatch theory, and sociological and economic literature empirically examining the validity of the spatial mismatch theory. Early Suburban Transit Demonstrations The federal government financed at least two suburban transit improvement initiatives before the summer 1965 Watts riots. A year after a public authority took over public transporta t ion service from p rivate operators in the St. Louis area in 1963, the United States Department of Housing and Urban Development financed a year-long demonstration of seven new express bus lines to the central business district from distant outer suburbs, as well as a new crosstown local line linking three of the same outer suburbs. The transit authority hired a consultant to document and evaluate the demonstration, which ran from May 1 964 to May 1965 (Gilman, 1 966). The consultant did not evaluate the new services from the standpoint of their impact on employment status but rather looked at ridersh ip, revenues, volume of serv ice, and o rigin and destination characteristics of riders during the course .of the demonstration. While nothing in the consultant's report mentions race i n consideration of service design or evaluation, the results still are interesting in that they show the crosstown route t o have had significant ridership. The seven express routes averaged revenue of 25. 1 cents per bus miles at the beginning of the experiment and a veraged 31 9 cents a year later, an i ncrease of 27 percent. The crosstown began with revenue of 10.1 cents per bus mile and c on cluded with 19.7 cents, an increase o f 95 percent. At the end of the demonstration, the crosstown carried more passengers per bus mile (.g1) than the average for the expresses (.85), but because many of its passengers transferred to or from other routes, and also because the expresses char ged a 10 cent supplement, each crosstown passenger brought in only half the revenue of passengers riding expresses. Few of the passengers riding expresses transferred. At the end of the demonstration a quarter of the crosstown riders were new, while the others benefitted from more direct service than the y previously had Forty percent o f express riders were new, but express buses serv ing districts with high levels of suburban employment had only half the ridersh ip per housing uni t served compared to express buses serving districts with no suburban employment. This suggests that as suburban employment increased, exp ress riding woul d decline. Also, while most express riders going to work traveled to jobs in the central business district, only one to two percent of crosstown bus riders going to work did so. Fifty to sixty percent of crosstown riders traveled to jobs along the crosstown route itself, whil e 30 to 2

PAGE 10

35 percent transferred to routes that took them to jobs in other suburban locations (Gilman, 1966, pp A 7 to A-10, 8-36 and 8-37, 50) These points suggest that the crosstown route was penetrating the suburban employmen t market, whi ch was growi ng, while the centra l business district express buses failed to do so. At the same t ime, the crosstown s revenue ge n eration was less than that of t he ex p res s buses, but given the high cost of central city expresses, with tHe i r high demand on equipment and l abor during the peak period compared to the base, compared to the much more even demand tor the crosstown se r vice, the crosstown's financial performance cou ld have e q ualled or bettered that of the expresses (the report did not touch on this po i nt). At the conclusion of the demonstration the transit authority continued operating live of the seven express routes as well as the crosstown. Another suburban demonstrat i on took place i n Memphis, testing the idea that bus serv ice inaugurated in subu rbs as they opened for occupancy would attract greater ridership than service started some t i me after the suburb became estab li shed. Several radial local routes serving the central bus i ness district were extended farther out to suburbs under construction. One route also served a suburban employment site w ith a few hundred worke rs. Results proved lackluster; r idersh ip t o the suburb a n employment s ite was a lmost nil/Memphis, 1965). Recent empirica l work on trans i t ridership patterns in Sacramento County, California, reaches conclusions similar to those of the St. Louis demonstrations. In Sacramento about half of a ll morning peak period transit trips terminate i n locations other than the centra l business d istrict or more broadly defi ned downtown. An even greater proport i on of trans i t trips made ove r an enti r e day terminate in suburban destinatio n s. All trips with the home as one end are defined as beginning at home (Thompson 199 7 a and 1997b}. Analysis of Watts R iots The first major study addressing the Watts ri ots, t he McCone Commission report (Go vernor's Comm is s ion 1965), identified ina deq u ate pub li c transporta t ion between the Watts ghetto and other parts of Los Angeles as a significant cause of social disc o ntent. Among the inadequacies that it id entified was t he absence of east-wes t bus service link in g Watts with suburban shop ping and employment centers. The repo r t recommended inauguration of east-west bus service as well as improve d north-south bus service t o downtown Los Ange le s and perhaps to southerly employment cente rs. It also recommended free transfers between all bus r outes in the a r ea. At the time passengers who wanted to transfer between routes of different agencies or private compan ie s o r even transfe r between routes that at one time were operated by different pr ivate compan ie s but in 1965 were operated by a public agency, had to pay additional full fares. Double-payin g was not too much of a hinderance to residents wanting to travel to downtown Los Ange l es, but i t was a major hinderance to those wanting to t r ave l to othe r destinations. The report came out at the same time as the creation of the U r ban Mass Transit Administration (Federal Transit Administration since 1993}, which took over incipient mass transit demonstration projects previously within the U .S. Department of Ho usi ng and Urban Development. As a consequence of the r eport, the State of California u s ed UMT A 3

PAGE 11

funding for an eastwest crosstown demonstration route linking Watts with other areas, as well as shared ride taxis, and other (what now are called! paratransit services to major suburban factories. The State of California Business and Transportation Agency later evaluated the service in a report, and while the study team could not locate a copy of it, the urban historian Sam Bass Warner summarized it as follows: After the Watts riots the transportation problems of the poor were made pla in, and the state of California, with federal funding, set up a demonstration pro ject in southeast Los Angeles to attempt to deal with the problem. A social science team intervened in behalf of the poor. . they discovered that the last bus might leave a factory's gates a few minutes before the men finished their daily shift; they d isc overed r o utes that could be instituted or altered to reflect the commuting habits of the residents ... The altering of existing pub lic transportation was not sufficient ... to this end the experiment rented bu ses for some new regular routes .. The cost of these services was high, but could have been much reduced if established permanently. The demonstration project was a success, but in 1971, when the federal funds were gone, predictably the grant by the government was not renewed and the s ervice was closed down. (Warner, 1972, pp. 143-144).' According to Warner, the report also conc luded that public transportation was important more for women and older people who lacked cars. Poor men, once they obtained a job with the benefi t of improved crosstown transit, could obtain the credit to buy a car, and they did so (Warner, 1972, p. 143). From this conclusion the belief became widespread that the suburban jobs-oriented transit demonstration failed. While it enabled some previously unemployed African-Americans to find jobs, with earnings from their new-found jobs, the African-Americans bought autos and stopped using public t ransit. Kerner Commission A couple of years later the Kerner Commission more broadly examined the phenomenon of urban ghetto rioting, which by then had occurred in several of the nation's larger cities. The Kerner Commission's transportation recommendations echoed those of the McCone Commission, though they were not given nearly as much importance in relation to other factors. In order to better connect central city ghetto residents with jobs that were rapidly suburbanizing, the Kerner Commission report recommended: The existing experimenta l mobility program, under the Man power Development and Training Act, should be greatly expanded, and should support. movement from one part of a metropolitan area to another. Aid to local public transportation. under the Mass Transportation Program should be similarly expanded on the basis of an existing experiment with subsidies for routes serving ghetto areas !Kerner, 1968, p. 418). 'Warner (1972, p. 143, n. 48) gives the following citation for the repo rt: State of California, Business and Transportation A gency Transportation-Employment Project, Interim Final Report, Los Angeles, January 1970, pp. 85-90. 4

PAGE 12

Spatial Mismatch Theory In 1968 John Kain (1968), an economist examining t h e link between A frican American unemployment, job suburbaniza tion, and center-city oriented mass transit systems, proposed what became known as the spatial mismatch theory. The spatial mismatch theory holds that the growing ratio between African-American and white unemployment rates results from urban job deconcentration coupled with housing discrimination against African-Americans. Poor transportation from center city ghettos to edge city jobs also figures in the theory. Job deconcentration takes low-skilled manufacturing and service jobs out of the centra l city to increasingly distant suburbs; housing discrimination keeps lowskilled African-American workers in central city ghettos. Increasingly costly commutes from the central city ghettos to edge city jobs depresses African-American employment The relevant statistic here is the ratio of African:Americantowhite unemployment, particularly for males, which I will call the racial unemployment ratio. Since the 1960s the racial unemployment ratio for a ll workers as well as for youth has increased in most urban areas of the United States to somewhere between 2 and 3 to 1 (Lichter 1988, Shulman 1986; Lerman 1986). Because the high rat io li kely contributes to hopelessness and crime, social scientists have sought to understand its persistence dur ing the era of desegrega t ion. Since the late 1 !:)60s the spatial mismatch theory has focused such research leading more recently to counter hypotheses. Because there is little disagreement about low-skill jobs moving to the periphery or about housing discrimination against African-Americans the spatial mismatch theory found ardent supporters. It a lso led to policy prescriptions. lnclusionary housing programs and demonstration projects for improved transportation links from ghettos to suburban employment have grown out of it (Hughs 1 987). Despite such appeal, the spatial mismatch theory has aroused skepticism. Taylor and Ong {1995) exam ined the mismatch hypothesis using journey-to-work data from the American Housing Survey for African-Americans and Hispanics, residing in central city ghettos, and suburban whites in the mids and mid-1980s. If the mismatch hypothesis were true, commutes for African-Ameri cans and Hispanics in the ghettos should be significantly longer than those for suburban whites, Taylo r and Ong argue. Their analysis of the data shows, on the contrary, that commute distances for African Americans and Hispanic whites are shorter than that of non-Hispanic whites and that relative commuting distances did not change between 1980 and 1990, a find ing also reported by Gordon, Richardson, and Jun (1991 ). Taylor and Ong conclude from this that minorities a re finding work close to horne and from this they discount the spatial mismatch theory. They observe that travel times for African-Americans and Hispanic whites were longer because of their greater reliance on public transportat ion. The Taylor and Ong findings that African-Americans work closer to horne than others could have another explanation. This is that the greater reliance of African Ameri cans on slow public transportation confines thern to a job search that is closer to home. Some African-Americans find jobs within their restrict ed search radius; many do not and are unemployed, thus accounting for the much higher unemployment status of African Americans compared to non -His panic whites. This explanation actually supports the spatial mismatch theory 5

PAGE 13

Others discount the spatial mismatch theory by arguing that transportation is a relatively small barrier to minority group employment compared to employer discrimination. Prior to the 1960s employers could pay African-Americans lower wages than comparable groups of whites. After the 1960s anti -discrimination laws prohibited this practice as well as discrimination in hiring, but wage discrimination was much easier to detect and prosecute than employment discrimination. Since that time, wages for African American employees have risen comparatively to those for whites (though still not to parity), but relative unemployment for African-Americans a ls o has increased (Shulman 1 986; Lichter 1988; Danziger and Gottschalk 1987)2 Two interpretations have been placed on these findings. Danziger and Gottschalk interpret these results as indicating that African-Americans are increasingly being assimilated into the white middle class suburban culture As African-Americans in ghettos obtain well-paying jobs, they move to the suburbs. Increasing jobl essness in the ghettos is caused because those who get jobs leave. The policy implication is that the ghetto should be the recipient of policy intervent ion. This finding can be true only if African-Americans who move to the suburbs have employment rates comparable to whites. If the spatial mismatch theory were true, the racia l unemployment ratio should approach unity for groups of African-Americans and whites living in the suburbs, controlling for skill levels, education and f amily structure. Research looking into this question found that parity does not exist. African American/white unemployment ratio is about the same regardless of the residentia l location of either African-Americans or whites (Ellwood 1986;.lichter 1988; Blackley 19901. Poorly-educated, lowskilled whites living in proximity to central city African American ghettos have only half the unemp l oyment rate of African-Americans in the same area; African-Americans li ving in the suburbs have double the unemployment rate of nearby whites having similar skill and education levels. Moreover, if transportation costs from ghettos to edge city jobs were to account for the 2 to 1 ratio, transportation costs would have to be many-fold higher than they already are. Some other cause is at the root of African-American unemployment (Ellwood 1986; Blackley 1990}. Shulman {1986}, who unlike Danziger and Gottschalk recognizes high African American unemployment in the suburbs, concludes that in the context of increasing nat iona l unemployment, firms discovered that hiring discrimination was profitable in all areas. As firms have increased wages for African-Americans, they have hired fewer African-Americans, particularly for low-skilled jobs. A possib l e rationale for firms relocating back offices and other activities requiring lowskilled labor to the suburbs i s to tap the large pool of white suburban housewives. Where does this leave us regarding the role that improved public transportat ion might play in increasing labor force participation? None of the l iterature explicitly examined the impact of transit linkages on labor force participation rates, but several findings touch on transp ortation policy measures. The finding that the rac ial unemployment ratio is rel atively unaffected by housing location with respect to employment suggests that impr oved transportation will have little impact on alleviating African-American or other unemployment. The proximity between jobs and housing is just not the iss ue here. This 'Lichter (1988} does not make this argument, but his tab l e 1 results confirm it. 6

PAGE 14

finding is consistent with that of unrelated research showing that urban structure, and particularly the proximity of housing and jobs, has little impact on travel behavior in general (Giul iano and Small 1 993). On the margin, though, transportation imp rovement could have some impact. Blackley (19901 and Ellwood (19861 both found that difficul t access between ghettos and employment centers could explain a small part of the rac ia l unemployment ratio. Blackley suggested that the employment rate of African-American females (but not males) could be improved marginally with better public t ran sportation links from ghettos to suburban jobs. Hughes (1987), who accepts spatial mismatch theory, recommended that based on assimilation theory from the Chicago School of Socio logy, improved t ransportation links would be preferable policy to inclusionary housing polic ies as a means for redressing the rac ial unemployment ratio. DADE COUNTY STUDY OF IMPORTANCE OF TRANSIT ACCESSIBILITY Hypothesis and Means for Testing None of the statistical analyses of the spatial mismatch theory directly addressed the question of whether improved public transportation cou ld affect labor force participation rates. They suggested that i f such a study were done, however, that it should not focus narrowly on linkages between a given area of residence and nearby jobs or distant jobs in one location such as a cluster of aerospace manufacturing plants The supply of jobs throughout an metropolitan area is what is important to people i n a g iven neighborhood who are looki ng for work. Thus, statistical analyses need to incorporate explanatory variables that measure how well all job s of a particular type (such as manufacturing or service) are linked to the neighborhood by public transportation; not jobs just in one area. None of the statistical analyses examined here have done this; neither have the demonstration projects. This research develops an accessibility measure indicating how well a part of a census tract (traffic analysis zonel is connected by public transportation to all jobs in the region. This measure then is added to other explanatory variables to explain the probability of a person of a particular type (race, gender, age groupl obtaining employment. If the coefficient for the accessibility measure is s igni ficant, the hypothesis is accepted that public transportation accessibility matters. Study Site Dade County, Florida is used as the site for this research, because it is home to large numbers of African-Americans and Hispanic white people, and it also is the only county in Florida with a well-developed t ran sit system. The Metro Dade Transi t Authority operates a la rge system of bus routes and one rail rapi d transit route within the county. Bus routes inc lude traditional l ocal and express radial r outes focused on the central business district; in addition, there are several crosstown local routes. Other local and express routes act as feeders to rapid tran sit stations. The rapid transit route (called Metroraill skirts the central business district at its mid-point, so it functions effectively as two r adi al central business districtor iented routes. Because the rapid transit line and some 7

PAGE 15

of the rad i al bus routes do not go through the cen t e r of the central business district, many CBDbound transit passengers get from their radia l routes to their final destinations by walking l engthy d istanc es or t r ans ferring to an elevated circul ato r minirapi d transit line in the central business district. Called the Metromover the mini rapid -transit service recently had two spurs added to it, doubling its length. Collectively the Metro Dade Transit Authority services account for a lmost half of all transit passenger trips in Flor ida and probably well over half of F lorid a transit passenger miles. System traffic a l so has been growing. In 1985 Metro Dade Tra nsi t Authority t raffic totalled 307 million passenger miles, distributed 267 mi llion to buses 1 m i llion to Metromover, and 39 m i llion to M etroral l. I n 1990 345 million passenger miles used the system, distributed 232 milli on to bu ses, 3 million t o Metromover. and 110 milli on to M etro rail {Mac Dorman & Associat es 1992 ; Cente r for Urban Transportation Resear c h 19961 In add ition pri vate owner-oper ators prov id e a large collection of jitney serv i ces in several areas of the county. I have no written infor mation on these, but j itney services are included In the Metro Dade transportat ion model. Their travel times. which are faster than fixed route t ravel times between a large numb er of pairs of traffic a n alysis zon es, are used in this s tudy elong with f i xed route travel times as further described below Study Design Following on the method of Ellwood's {19861 analysis of t h e impact of accessibility on A f rican-Ame ri can unemployment in Chicago census tracts, this study postulates un employment rates in a sample of traffic analysis zones of a larg e metropolitan area {in this case, Dade County Florida) as a function of race, education level and accessibility. Th e study differs from Ellwood's in its definition o f access ibil ity measures and in usin g traffic analys is zones rather than census tracts as the unit of analys i s. It a l so differs by adding binom ial logit analysis of the variable s in addition to ordinary least squares analy sis, whi l e It a l so l ooks at the influ ence of tran s i t accessibility on transit mode split for African Americans, Hispanic whites, a nd nonH ispanic Whites, and finally it looks at the relationship between transit access ibility for all three groups and automobile ownersh ip rates. Sample Size The data set that I u sed includes all traffic analysis zones with 80 percent or more of its population being African-American s (about 90 traffic analysis zones). addi tional traffic analysis zones east and south of the central business district, and in addition every tenth traffic ana!l,isi s zone not otherwise selected. Such considerat ions yielded a sa mpl e siz e of 23 7 traffic analysis zones, and for each of these I calcu l ated the three accessibility indices as defined by equation 1 Access ibility Indices Ellwood {1986) used three measures of accessibility: jo bs within a thirty minute transit ride of a neighborhood the ratio of jobs-to-people within each neighbo r hood, and the average navel time to a job from a g iven neighborhood. He rejected accessibili t y 8

PAGE 16

measures for each census tract. using planning-department-defined neighborhoods instead. because he did not have the means to calculate finer measures, and he considered a census tract too small to be a viable labor market. None of his accessibility measures had any statistical impact on employment rates for ce n sus tracts. Unlike Ellwood, I calcu l ate t ran sit accessibility measures for in dividual traffic ana lysis zones, which are smaller than census tracts. The output of t ransit mode ls now in use by most metropolitan planning organizatio!'s inc lud ing the Metro Dade County MPO. makes TAZ transit accessibility measures calculable. My study ques tion, whether transit connections to jobs throughout the region makes any different in emp loym ent status, also demands a precise measure. This is because transit users w i ll not walk very far to a transit stop--typically no more than a quarter mile. This means that adjacent traffic analysis zones could have very different transit connectivity to the region, which could give their residents very different employment opportunities. The Metro Dade County MPO transportation model output inc lu des transit door-to door travel times, including walk and transfer times, between all pairs of traffic analysis zones in the county. From such data, I used typical gravity model measur es for t ransi t accessibility: A & = L TT1Mij2 JOBSP< j Equat ion 1 wher e A;, is the transit accessibility from traffic analysis zone i to jobs of type k in all of the region, n is the number of traffi c analysis zones in the regio n, TTIM is doo r to-do or transit trave l t ime in the morning peak period from traffic analysis zone i to traffic analysis zone j, and JOBSiJ< are the number of jobs of type k lo cated in traffic analysis zone j. I used three types of employment: industrial service, and commercial, which sum to total employment. Each traffic analysis zone used in the study thus has thre e transit accessibility indices. one for each type of employment. These are in dicated as A IND, A_COMM. and A_SER in t h e rest o f the study. This approach differs from earlier work that I conducted on Sacramento, California transit data, where I extracted gravity-like accessibility measures from an estimated direct deman d model of transit in the area. That study showed that while h i gher density suburban employment would attract transit ridersh ip, lower density employment would do so at a far lower rate. I chose a different measure for this study because I explicitly am examining the hypothesis that transit linkages to all employment makes a difference. Miami/Dade County Metropolitan Planning O rgani zation made avai l able to this study their input data to t heir 1990 a l ternative A TRANPLAN transportation simu l ation mode l. A student Transit Fellow. Kevin Til bury, ran this to obtain transit tra ve l t imes between a ll pairs of Dade County traffic analysis zones during the morning peak period for use in equation 1. The model yields travel times by several compet i ng transit modes between all pairs of traffic analysis zones. I was interested in only those alternatives that did not rely on auto access: jitney, all local bus, and various combinations of loca l bus and either 9

PAGE 17

express bus or r ail. F or this reason, I defined TTIM;; as the shortest transit t r ave l time from each of the three transit poss i bi l ities be t ween each i and j I did not consider fares in the accessibility calculati ons. Extracting trave l times from intermediate files of the model s output proved diffi cult. Ultimately, it proved feasible to do so only on the basis of one traffic analysis zone at a time. I could select any traffic analysis zone i and after about a minute of fortran calculat i ons have the three accessibility indices shows i n equation 1 calculated for that traffic ana l ysis zone. Table 1 presents summary data statistics for the three measures for the 237 zones used in the study. Other Variables and Weighting The 1990 Census Transportation Planning Package (CTPP) for Dade County provided the other variables used in the study Table 1 defines these and presents summary statistics for the 237 zone sample Unfortunately, the Census Transportation Planning Package offers only rudimentary socio-economic variables. It would have been des i rable, for example, to have unemp loyment rates, median incomes, and auto ownership separately for Afri can-Americans, Hispanic whites, and non Hispanic whites for each zone, but the CTPP provides unemployment rates on l y for males and females and median incomes and a uto ownership aggregated over all races and households in each zone. Use of the Public Use Microsample (PUMS Data) from the census would overcome this limitation, but PUMS data are stripped of geographic location except at gross levels, making imposs i b l e the definition of accessibility indices which is the heart of this study. So CTPP data were used. To partially compensate for the absence of unemployment rates f or each ethnic group, I ra n analyses of the same data three times, each time weighted by the population o f the ethnic group in whose behavior I was investigating The first three variables in table 1 are the transit accessibility variables Each of these exhibits a high degree of variance showing that transit accessibilities vary widely over the sample of traffic analysis zones. Although I developed three access i bility measures in the belief that transit accessibility to different types of work might differ widely, the three measures are highly collinear for the 237 zone sample: the correlation coefficient between A IND and A SER is .8658, that between A IND and A COMM is -.9401, and that between A_COMM and A_SER is .9703. In the analyses that follow, I use A COMM as an index represen t ing transit accessibility to all jobs. M E DHHINC is median household income, and it a l so has a high degree of variance. I include it in analysis of automobile ownersh ip, VEH HH. Income is often presented as the most s ignificant explanatory variable for auto ownership, and auto owner ship often is used as a proxy variable for income because the two variables are often so highly corre l ated. Despite such relationships, for the 237 zone sample used here the correlation coefficient between MEDHHINC and VEH_INC is only .5885, moderate but not high correlation The relatively low correlation may represent a saturated auto market; the var i ance for VEH _HH relative to its mean is a third lower than that for MEDHHJNC, probab l y reflecting a re l atively high degree of auto ownership in poorer households. TOT POP i s tota l population in each traffic analys i s zone, while PAFAM, PHW, and PNHW represent the proportion of each tract's populat ion that is African-American, Hispanic white, and non-Hispanic white. The relat i vely high mean for PAFAM reflects the 10

PAGE 18

choice of sample inclu ding all traffic analysis zones in Dade County with 80 percent or more African-American population, about 91 zones. The correlation coefficient between PAFAM and PHW is -.7720, between PAFAM and PNHW is -.6137, and between PHW and PNHW is -.0256 for the 237 zone sample, indicat ing a moderate amount of segregation between and Hispanic white and non Hispanic white populations, but no segregation between Hispanic white and non Hispanic white populations. P12_17NE is the proportion of youngsters between 12 and 17 years of age who are not enrolled in school. Ellwood ( 1986) and Blackley ( 1990) found that variables reflecting education and family status {proportion of families with a single mother) were significant predictors of employment status. Areas with an education tendency, ind icated by a high proportion of school-age children being i n school, had much lower unemployment. Areas with a two parent househo l d head also had lower unemployment. From the CTPP I could extract a variable reflecting education status but not one reflecting family status. PM_ UM and PF U are the proportions of males and females respectively, seeking work but who cannot find it. Finally, TMS_A FAM, TMS_HW, and TMS _NHW are the proportions of African-American, Hispanic white, and non Hispanic white workers in each zone who use any type of transit to get to work. Results Table 2 presents the results of an ordinary least squares (OLSI analysis of variables thought to be responsible for employment similar to that Ellwood (1986} presented in his table 4.5. The major difference is that I used three weightings for the ordinary least squares, one for each of the three racialfethnic groups under study. Another is that I introduced vehicles per household (VEH_ HHI as an explanatory variable, because it would be difficult to interpret the impact of transit accessibility unless auto ownership were controlled for. One would expect transit accessibility to have no explanatory power in areas with high auto ownership. Results for all three weigtl!ings are simila r H igher auto ownership rates are highly related to lower unempl oyment, and the relationship is significantly stronger for African Americans (particularly African-American females} than i t is for either Hispanic or non Hispanic whites of either gender. These resu lts are consistent with t he transit service demonstration in the wake of the Watts riot, suggesting that those with income buy autos, and that employment is a greater threshold for auto ownership for African-Americans, particularly females, than it is for other groups. These results also could suggest that the higher the auto ownership In a zone, the greater the likel ihoo d of employment and that auto ownership is particularly important for the employment of African-American females. Similar to Ellwood's findings, tabl e 2 results suggest that race or ethnicity a lso affects unemployment rates. I n zones that are p redo minately African-American, the inclusion of either H ispanic or non-Hispanic white s significant l y reduces the rate of unemployment for males, but not for fema les In zones that are predominant l y H ispanic white, the in clusion of African-Americans significantly increases unemployment. while the inclusion of non-Hispanic whites significantly decreases unem ployment. In zones that are 1 1

PAGE 19

predominantly non-Hispanic white, the inclusion of African-Americ a ns or Hispanic whites Increases unemployment, although the results are not s i gnificant for Hispanic white males. The v ariab le for education (the percentage of t 2 t 7 year -olds not enrolled in school) is somewhat important, though importance varies by ethnic groups and gender. For all groups, there is a tendency for zones that are more education-oriented, as indicated by this variable, to have lower unemp loyment but the result is highly s igni ficant for Hispanic white an d non-Hispanic white fema les It is not quite sign i f icant at the five percent l evel for Hispanic w hite and non-Hispanic white ma l es, it i s somewhat less significant for African-American female s, and It is insignificant for African-Am erican males. As Ellwood found, transit accessibility generally has insignificant power in explain ing unemployment l evels, but where it does have some significance, its sign is contrary to a priori expectations. As discussed further below with in sight from the remainder of the tables, these results tend to show that transit accessibility is not important for increasing employment l evels, except possibly for African:Americans to a limited extent. H owever, transit accessi bility is important to the well-being of all three groups in other ways, as discussed below. First, we need t o consider whether tabl e 2's r esults ar e valid. It i s shown so that results from Dad e County can be com par e d with Ellwood's results. Despite the c l ose compa rison, the table 2 results are suspect, becaus e the models f rom which they derived are likely misspecified. The dependent variables are proportions whose potential ranges are from 0 to 1, and OLS is generally considered a n i nappropriate explanation for such a dependent v ariable. A more suitable specific ation for such dependent variables i s the binomial logit model, which in this case can predict the probability of two conditions: unemployment or employment. The two p robabilities sum to one. The form of the model estimated here is: m explb0 + L b, X,) PMIUM) = ---..!':.:.''.,..-.. 1 +-explb0+ E b,. X,) .. Equation 2 where PMIUM) is the probab ility of male u n employ ment, b0 is a constan t to be estimated, and each b, is a coefficient to be estim ated for each e xplanatory variable X,. There ere m explanatory variab les. Table 3 presents results of the logit esti mations of unemployment probabilities for males and females for each of the three racial / ethl)ic groups using the same explanatory variables and weighting systems as in tabl e 2. The Chi square statistics show that each of the models has s ignif icant explanatory power, but none of the coefficients tor the exp l anatory variables is significant at the five percent l e vel. The s ign s of the variables, however, are in the same direction as those in table 2. I n general table 3 s hows the same tendencies of variabl e relations as t able 2 but none of the relations are statistically 12

PAGE 20

significant or even close to statistically significant. There is no evidence from table 3 that transit accessibility affects unemployment for either gender for either racialfethnic group. Table 4 examines the hypothesis that transit accessibility is important to transit ridership. Here the dependent variable is work trip transit mode split for each racial/ethnic group in each traffic analysis zone used in the study. Again the variable is in the form of a proportion that can vary from 0 to 1. By specifying the dependent variable as a function of a binomial logit relation shown in equation 2, I estimate the probability of a work trip be ing undertaken by transit compared to the probability of it being undertaken by all other modes. Again the Chi-squared statistics indicate that a ll of the models shown in table 4 have explanatory power, but none of the individual explanatory variables are significant at the five percent level. The exp l anatory variab le that comes closest to the five percent level is transit access ibility to commercial employment in the model exp laining transit mode split tor African-Americans. This variable's coeffi cient has a 94.2 percent probability of being drawn from a distribution whose mean is greater than zero. Transit accessibility to commercial employment has much less explanatory power for non-Hispanic white work trip transit modes spirt and even less for Hispanic white work trip transit mode split, but these variables still go farther than any others !except the constants) in explaining transit mode split. Table 5 shows results for expla ining automobile ownership using ordinary least squares analysis: Here results are very strong. Median income is highly significant for all three racialfethnic groups, as is transi t accessibility. Higher transit accessibility is associated with lower auto ownership. The impact here is particularly strong for Hispanic and non Hispanic whites. Et hnic composition also appears impo rtant, all else equal; Hipanic whites have a higher propensity to own autos than Afro-Americans or non-Hispanic whites. F in ally, the unemployment rate for African American females has a highly significant negative impact on auto ownership in tracts that are populated primarily by African-Americans. Table 6 also explains automobile ownership, but it uses Poisson regression to do so. 1 have included table 6, because ordinary least squares !OLS) may not be an appropriate model for explaining a dependent variable whose lower bound is 0 Poisson regression is suitable for such a variable, but for Poisson regression the dependent variable must be in the form of a count variable. The model estimated is: "' VEHIHH) = explb0+L b, X,l. , Equation 3 where VEH!HH) are vehicles per household and the explanatory variables are as given i n equation 2 To estimate equation 3 with Poisson regression, the dependent variable VEH_HH must be in the form or a count or in teger variable. Ideally, this type of equation should be estimated wrth individu a l records taken from the U.S. Census Public Use Microsample, but 13

PAGE 21

again because such records have locat ion stripped from them, that course could not be followed in a study where the Importance of accessibility is being investigated. As a second best course of action, I recoded VEH_HH to an integer variable: traffic ana lysis zones with 0 to 0.499 autos per household were assigned 0 autos per househo ld, 0.500 to 1.499 autos per household were assigned 1 vehicle per household, 1.500 to 2.499 were assigned 2, 2.500 to 3.499 were assigned 3, and 3.500 to 4.499 were assigned 4. The recoding created a new dependent variabl e VEH_HH. This is an integer variable that ranges from 0 to 4 Its mean is 1.4895 and its standard deviation is 0.6549 over the 237 cases in the sample. Table 6 presents the results from the Poisson regression Results generally are similar to those for table 5, though they are not as strong, and fewer ex planatory variables have significant coefficients In column 1, the mode l results with AfricanAmerican weighting, transit accessibility is not statistically significant for explaining auto ownership, though household income is. None of the other explanatory variables are significant. In column 2, the model with Hispanic white weighting, transit accessibility is an important depressant for auto ownership; income is not statistically important, and the remaining explanatory variables are even less significant. In column 3, the model with non-H ispa nic white weighting, income and Hispanic white presences are highly significant, and transit accessibility is also significant. This means that non-Hispanic white zo nes that have low income but high t ransi t accessibility, auto ownership is depressed. Inc reas ing the proportion of Hispanic whites in such zones increases auto ownership, suggesting a cultural affinity of Hispanic whites for car ownerhip, all else being equal. Even so, results from column 2 suggest that Hispanic whites may reduce car ownership if theyh l ive where there is high transit accessibility. D ISSCUSSION AND CONCLUSIONS On first impression, tables 2 and 3 seem to disprove the hypothesis that transit accessibility decreases unemployment for African-Americans and Hispanic whites. To the extent that there is any statistical relationship, it Is in the wrong direction. Despite first impressions. however, tables 4 and 5 or 6 offer additional information that must be considered before making conclusions. The positive relationship between accessibility and unemployment shown in tables 2 and 3 suggests somehow that the prov ision of high transit accessibility increases unemp loyment, at least for African-Americans. A possible way that transit service could do this is that households with l imited numbers of automobiles and higher levels of !Jnemployment locate in areas with high transit accessibility, in order to increase their own mobility. This explanation is consisten t with tables 2 and 3. as well as table s 5 or 6. Tables 5 or 6 generally show a very strong inverse r elat ionship between automobile ownership and transit accessibility (except for African-Americans in the case of table 6). Some of the auto-limited households that located to take advantage of transit accessibility, particularly African-American households, may h ave members who use transit to go to work, consistent with t able 4. Other auto-limited households may send their employed members, if any, to work in the few cars available. All such households may find transit's principle mobility advantage for nonwork trips. 14

PAGE 22

There is another possible explanation for the weak l ink in the wrong direction between unemployment and transit accessibility. This may be that transi t agenc i es supply more transit service to areas w ith high unemployment. That is, transit agencies and/or j itney operators may conscious ly provide more ser vice to areas that they perceive a s auto deprived, consistent with tables 5 or 6. Such areas may have higher levels of unemployment. The services may not make much of an impact on unemployment, though the fact that African-Americans appear to take advantage of hig her leve ls of transit access ibility to travel more to work by trans it, as indicated by table 4, suggests that it may have some i mportance for a lleviat ing unemployment. Desp i te that, the more important impact of the services may be for non-work t rips. Both explanations are consistent in conclud i ng that trans i t accessibility to the rest of the region is more important for non-work trips than work trips. Consistent with results from Black ley (1990) and Ellwood (1986), this study shows at best weak links between region -wide transit accessibility and work trips, but it also shows very s.trong, inve rse links between transit access ibility and auto ownership. This latter l inkage suggests t ha t transit accessibility has to have importa nce to some aspect of mobility, and the non-work trip is left by default. I found, but did not comment on, similar resu lts from work i n Sacramento In Thompson 11997a) the ana lys is was based upon transit 24 hour week day travelers and transit mid-day serv i ce leve ls. I n Thompson (1997b) the analysis was based on morning peak period transit travel (mostl y work trips} and service levels from the morning peak period (as in this study) In the Sacramento work, the link between transit accessibility and transit usage was found to be much stronger for the ali -day analysis ( 1997 a) than for the peak period analysis 11997b). More work needs to be done here to determine how transit service is provided in Dade County, by both regular and jitney operators. More work a lso cou l d be done on the accessibility measures. Transit fares could be incl uded; measures a l so could be defined to isolat e the effect of jitney services as well as the ac cessibi lity of the centra l bus in ess district compared to s ubu rban jobs. In general, it can be concluded tentatively t hat: 1. Transit accessibility is important in explaining African-American transit travel to work. That is, the more broadly is cast the transit net from a particular traffic analysis zone, the more likely it is that AfricanAmericans will use that net to trave l to work. Transit accessibility, though, does n 0 t appear to have an impact of reducing unemployment for African -Americans or other groups. 2. T ransit accessibility is important to Hispanic white and non-Hispanic whites p rimar ily for non -work trip mobility. Lower income members of both groups appear to use high transit accessibility to lower the number of autos In their households. Firmer conc lusion s must await investigat ion into how changes t o the defintion of accessibility affect outcomes (such as using employment density in the measure), as well as investigation into how transit agencies and other serv ice p r oviders decide where to place transit service. 15

PAGE 23

REFERENCES B l ack l ey, Pau l R. 1990. Spatial mismatch in urban l abor markets: Evidence from large U.S. metropoli tan areas. Social Sciences Quarterly 7 1:39-52. Center for Urban Transportat ion Research 1996. Performance of Florida's Public Transit Systems, Executive Summary. Tallahassee: F l orida Depa rtment of Transportation, Public Transit Office. Danziger, She l don and Peter Gottschalk 1987. Earni ngs inequa lity, the spatial concentration of poverty, and the underc l ass. Amer ican Economic Review 77: 211-215. Ellwood, David T. 1986. The spatial mismatch hypothesis: are there teenage jobs missing in the ghetto? i n The Black Youth Employment Crisis, Richard B Freeman and Harry J Hol zer, editors Chicago : Uni ve rsity of Chicago Press: 147-185. Giuliano, Genevieve and Kenneth A. Small. 1993. Is the j ourney to work exp l ained by urban structure? Urban Studie s 30: 1485-1501. Gilman, W C. & Company, Inc., Transit Consu l tants. 1966. The Radial Express and Suburban Crosstown Bus Rider: A Mass Transportation Demonstration Project, Report /NT_MTD 8 Washington, D C : U.S Department of Housing and Urban Development. Governor's Commission on the L os Angeles Riots. 1965. "Violence in the city: An end or a beginning?" In Robert M. Foge l son, ed. { 1 969}. The Los Angeles Riots. New York: Arno Press: Pp. i -xix and 1 109. Gordon. Peter, Harry Richardso n and M Jun. 1991. The commuting parado x : Evidence from the top twenty. Journal of the American Planning Association 57:41 6'420. Hughes, Mark A l an. 1987. Movi ng up and movi ng out: confus i ng e nds and means about ghetto dispe r sal. Urban Studies 24:503-517. Kai n, John F. 1968. Housing segregation, Negro employment and metropolitan decentralizat io n. Quarterly Journal of Economics 82:175-197. Kerner, Otto. 1968. Report of the National Advisory Commis s ion on Civil Disorders. New Y ork: Bantam Books. Lerm a n R obert 1. 1986. Unemployment among low-income a nd b l ac k y outh: A review o f c a u ses programs, and policies. Youth and Society 17: 237-266. Lichte r Dan i el T. 1988. Racial differences in underemployment i n American cities. American Journal o f Sociology 93: 771-792. 16

PAGE 24

MacDorman & Associates. 1992. To Classify Transit Services: Eight Case Studies : Miami Metropolitan Area Summary. Washington, D.C.: U.S. Department of Transportation, Federal Transit Administration : 85-103. Memphis Transit Authority. 1965. Mass Transportation Studies in Memphis, Transit . System's History 1956-1965, Suburban Ridership Demonstration Project. Washington, D.C.: U S. Housing and Home Finance Agency. Shu lman Steven. 1986. D iscrimination, human capital, and black-white unemployment: Evidence from cities. The Journal of Human Resources XXII: 361 -376. Taylor, Brian D. and Paul M. Ong. 1995. Spatia l mismatch or automobile mismatch? An explanation o f race, residence and commuting in US metropolitan areas. Urban Studies 32: 1453-1474. Thompson, Gregory l. 19g7a "Achieving Suburban Transit Potential: Sacramento Revisited." Transportation Research Record, forthcoming Thompson, Gregory l. 1997b. "The Equity Imp lica t ions of Light Rail Investment: The Sacramento Case, Department of Urban and Regional Planning paper series, Florida State University Warner, Sam Bass, Jr .. 1972. The Urban Wilderness. New York: Harper and Row. 17

PAGE 25

TABLES Tab le 1. Table 2. Table 3. Table 4. Table 5. Table 6. Means and Standard Deviations of Variables Used in Regressions Regression Results OLS Weighted as Indicated Logit Results Weighted as I ndicated Transit Mode Split Results Explaining Auto Ownership OLS Analysis Explaining Auto Ownership Poisson Regression Ana lys is (Dependent variable VEH_HH has been recorded to 0, 1 2, 3, 4) 18

PAGE 26

Table 1. Means and Standard Deviation s of Variables Used in Regression s Variable O&Scription Mean Standard Deviation Cases from FSUTMS model o f M etro Dade Transit and jitney service, alternativ e 90A A_IND A_COMM A_SE R transit accessibility to industrial e mployment transit accessibi6ty to cxmvnercial employment transtt accessibifity to service employment 35. 30 72. 69 196 98 from censvs transporlltion planning package, Dade COUnty, 1990 MEDHHINC median household lnoome $38.044 VEH_HH vehicles per house hold 1.40 TOT_POP total population 2,673 PAFAM proportion African-America n 0 .44 PNH W p roportion nonh ispanic white 0.22 PHW proportion hispanic white 0.30 P12_17NE proportio n of 12 to 1 7 year old not i n school 0 08 PM_UM proportion of males in work force unemployed 0.09 PF_U proportion of females in work force unemployed 0 .11 TMS_B transH mode split-African American 0.12 TMS_W transit mode split-no n hispanic white 0 04 TMS H transit mode split-hispan ic white 0.05 23.20 46. 27 142 .72 $22,959 0.58 2 086 0 42 0 26 0.31 0 10 0,07 0 10 0 .14 0 13 0 .10 237 237 237 237 237 237 237 237 237 237 237 237 237 237 237

PAGE 27

Tabl e 2. Regression Results OLS weighted as indicated Independent Variable Constant VEH_HH PAFAM PHW PNHW P12_17NE A_COMM n F adjusted r squared ...... .... . PM_UM PF _U PM_UM PF U PM_UM PF U (1) (2) (3} -. ( 4 ) (51__,,,, (6) 0 1801 -0 2129 0 1080 0.1348 0 0511 0.0616 (9.513) (10.260) (6 974) (7.941) (4 558) ( 5 .0 74 ) -0.0477 -0.0740 -0. 0227 w -0.0324 -0 0185 -0.0267 (-4.795) (-6.7 91) (-3.429) ( -4.455) (-3.253) ( -4.333 ) 0 .0641 0 0395 0 1 131 0 .1004 (5 1 96 ) (2.922) (9 939) (8 1 56) .0597 -0. 0072 0.0514 0.0597 (-2 204) (. 242) (4 920) (5.280) -0. 0764 -0.0071 -0 0538 .. .064 9 (-2.340) ( 941) (-4. 119) (5 929) 0 0116 0.0686 0 0553 0.0349 0 0448 0.1 11 6 (.227) (1.2 32) (1 738) (3.287} (1. 775) (4.085) 0 0003 0 0002 0.0000 0.0002 0 .0001 0.0001 (2 830) (1. 530) (0 372) (1.997} ( 739) ( 1.581) 237 237 237 237 237 237 25. 2 26.7 23.7 43 7 31.47 36.13 0.34 0.35 0 .32 0.47 0 39 0.43 (tstatis lics in parentheses)

PAGE 28

Table 3 Log it Results-Weighted as indicated Independent Variable ................................... De enden t Variable ............................. African-American Weighting Hispanic hlte Weighting Non ispa nic White Weighting PM UM PF U PM UM PF U PM UM PF U ----(1) (21 __ (3) ( 4)__ (5) ( 6 ) Constant -1. 4688 (1.4 72) VEH_HH -0 4181 (-0 765) PAFAM PHW -0.6049 (-0.355) PNHW -0.9478 (-0 417) P12 _17 NE 0 1569 (0. 058) A_COMM 0 .0021 c (0. 4361_ n 237 Chi -squared 25.8 s ignifi cance l e vel 0.00 (z-slallstlcs I n paren theses ) -1. 1934 (-1. 218) -0.6568 (-1.201) 0 0732 (-0.047) -0.8186 (-0.375) 0 6394 (0. 241) 0.0012 (0.247) 237 43.5 0.00 -1. 9550 (-1. 077) (-0.443) 0 6903 (0.583) -0 9515 (-0 538) 0 6856 (0. 198) -0 0001 (-0 007) 237 87.0 0.00 -1.6513 ( -0.989) -0.4140 (-0.567) 0.3618 (0.317) -1.3691 (-0.800) 1.0918 (0. 357) 0.0010 (0. 139 ) 237 86 5 0.00 -2 8554 (-2 012) 3894 (-0.526) 1 7015 (1.356) 1.07 1 2 (0.771) 0 8270 (0. 265) -0 .0001 (-0 013) 237 106.4 0 .00 2 6S85 (2.009 ) -0.4966 (-0.713) 1.4788 (1 .205) 1. 1 305 (0.864) 1.6459 (.597) 0 0005 (0.063 ) 237 1 15.7 0.00

PAGE 29

Tabl e 4. Transit M ode Spl it Results In dependent Variables Constant VEH_HH MEDHINC P12_17NE PM_UM PF U A COMM wei ghti n g n Chf-squared Significance level (z -statistics in parentheses) fransit Mode Split TMS AFAM T MS_HW TMS_NHW {1} 12) (3) 1 643 -2.202 -3. 19 1 ( 1.247) ( 1.021) (-1 693) -0.018 -1.0 47 -0.541 (.021) (-0.885) (.493) -0.000 -0. 000 0 000 (-0 752) (-0.069) (-0.120) 0.387 0 509 1.466 (0. 1 51) (0. 122) (.393) -0 745 4.214 -0. 033 (-0 230) (0.545) (-0 003) 0 677 -0.589 1 .751 (0.299) (-0 076) (-0 171) 0.009 0.007 0 012 j1.690l Afro-American jO 68!) Hispanic White p 165) N on-Hispanic population populati on white popu l ation 237 237 237 43.300 38.409 66. 2 1 6 0 .000 0.000 0.000

PAGE 30

Tab l e 5. Explaining Aulo OWnership OLS Analysis Independent Variable VEH_I-IH VEH_HH VEH_Hi'l (1) ( 2 ) (3) Constant 0 61281 1 50980 0 93832 .. (6. 213) ( 15.833 ) (11 348) A_COMM 00221 0.00470 0 00558 ( -5.911) (-10.603) ( 9.240) P 12 17NE 0 04178 0.48244 0 .054 11 (0. 1 97) (-1.949) (0 241) PM_UM -0 .14298 -0. 33188 0 46381 (-0 503) ( 0 612) (-0 794) PF_U -1.01630 -0 .59 0 1 3 -0. 97984 (-4. 089) ( 1.185) (1 825) PHW 0 35297 1 02820 -(3. 083) (11. 738) PNHW -0. 23938 -0.85305 (-1. 674) ( 7 523) PAFAM -0. 23800 ( -2.3 76) MEDHHINC 0 0000 3 0 00002 ( 15 .581l ( 11 1 45) weighting African Ameri can Hispanic Non-HISpanic population while population white population n 237 000 237. 000 237.000 F-statistic 113. 990 89. 100 69. 040 Significance level 0 000 0 000 0 .000 Adjusted r -squared 0 770 0 723 0 .669

PAGE 31

Table 6 Expla ining Auto Owners h ip Poi sson Regress ion Analysis (Dependent variable VEH_HH has been recoded to 0 1 2 3, 4) Independent Var iable Constant A COMM P12_ 17NE PM_ UM PF_ U PHW PNHW PAFAM VEH_HH VEH_HH VEH_HH (1) (2) (3) 0.01562 (0.044) -0.00225 (-1.442) 0 12394 (0 148) 0 74533 ( 0.638) ..0.62805 (..0. 595) 0 .22363 (0.540) ..0.22891 (..0.449) 0 .54459 ( 1.903) -0.00464 ( 3 070) 0.49428 (-0 589) ..0. 07007 ( 0 .039) 0.71270 (0.44 5) -0.48558 (-1. 320) 0.04347 (0.180) ..0.00453 (-2.453) ..0.02974 ( 042) ..0.66798 (..0. 397) 0 07313 (..0.048) 0 .79302 (3 163) 0.24077 0.62926 (-0 720) (1. 891) MEOHHINC 0.00002 0 .00001 0.00001 .. (2.73 1 ) (1.857) (2.87 4 ) weighting Africa nAmer i can Hispanic Non-Hi spanic n Chi -squared Signi ficance le vel R_p popula ti on white population white populat ion 237.000 237.000 237.000 75.719 24.559 24.750 0.000 0.001 0 .001 0 .547 0.610 0.572