JOURNAL OF FINANCIAL AND OUANTITATIVE ANALYSIS Vol. 43. No. 1, March 2008. pp . 245-266COPYRIGHT 2008. MICHAEL G. FOSTER SCHOOL OF BUSINESS. UNIVERSITY OF WASHINGTON. SEATTLE. WA 98195

The Genesis of Home Bias? The Location andPortfolio Choices of Investment CompanyStart-Ups

Jerry T. Parwada*

Abstract

Fund managers' bias toward geographically proximate securities is a well-researched phe-nomenon, yet the origins of managers' location choices have received little empiricalscrutiny. This paper traces the employment and geographic heritage of 358 entrepreneurialfund managers and analyzes the determinants of where they locate their firms and stock se-lections. The evidence suggests that start-ups tend to be based close to the origins of theirfounders and in regions with more investment management firms, banking establishments,and large institutional money managers. New money managers show a strong local bias intheir equity holdings, three times the levels previously documented for mutual funds. Thepropensity to invest closer to home correlates strongly with the presence of sub-advisoryopportunities from institutional investors in the vicinity. While home bias levels betweenmanagers who relocate with their start-ups and the rest of the entrepreneurs are similar,preferences for stocks that were formally local persist.

I. Introduction

A growing literature documents strong links between investors' decisionsand geographic location beginning with French and Poterba (1991). The phe-nomenon is reported for individual investors (Lewis (1995), (1999), Huberman(2001), and Grinblatt and Keloharju (2001)) and important intermediaries suchas mutual funds (Coval and Moskowitz (1999), (2001), Hong, Kubik, and Stein(2005), and Ivkovic and Weisbenner (2005)). The role of geography in economicactivities has continued unabated despite phenomenal strides in information tech-nology and the attendant reduction in transportation and communication costs (seeCoval and Moskowitz (1999), p. 2047). Given the apparent economic benefits toinvestors biasing their portfolios toward their home base (e.g., Hau (2001)), it issurprising that very little is understood about how intermediaries choose where

* Parwada, j.parwada@unsw.edu.au, School of Banking and Finance, University of New SouthWales, tJNSW Sydney, NSW 2052, Australia. I gratefully acknowledge financial support from theUniversity of New South Wales, Faculty of Commerce and Economics, t especially thank JoshuaCoval (the referee) and Hendrik Bessembinder (the editor) whose incisive comments improved thestudy considerably. For helpful discussions and comments, I am grateful to David Gallagher, StefanRuenzi, Andrei Simonov, Peter Vassallo, Donald Winchester, and participants at the European FinanceAssociation 2005 meeting. Maria Guo provided exceptional research assistance.

245

246 Journal of Financial and Quantitative Analysis

to locate in the first instance. Moreover, virtually no research has targeted rela-tions between the characteristics of investment companies' location choices andthe geographic spread of their portfolio holdings. In this paper, I attempt to fillthis gap by examining the location decisions of fund managers who start their owninvestment management firms. I also examine how fund managers' stock hold-ings relate to their former and newly chosen locations. Specifically, I address twomain research questions: i) which location characteristics attract investment man-agement start-ups?; and ii) do entrepreneurs tilt their portfolio holdings towardstocks that are close to where they locate?

Coverage of fund manager entrepreneurship in the popular press implies thatcertain locations encourage start-up activity more than others. New investmentadvisory firm formation in Boston, for example, is often credited to the influenceof mutual fund giants located in the city such as Fidelity Investments and Van-guard.' I examine the selection of locations by money management start-ups as adiscrete choice problem where founders' hypothesized geographic preferences aremotivated from the well-established industrial location theoretical literature (e.g.,Dumais, Ellison, and Glaeser (2002) and Figueiredo, Guimaraes, and Woodward(2002)). I define entrepreneurial fund managers' home bases as their last placesof employment and examine whether the location of start-ups relates to the char-acteristics of the places hosting their former employers. I then apply the Covaland Moskowitz (2001) measure of local bias in mutual fund portfolio holdings tothe stocks held by start-ups in their first year of investment operations to performtests of local bias in portfolios of start-ups.

To answer the questions raised in this paper, I use a unique hand-collecteddata set of over 350 seasoned fund managers who start their own firms. I findthat start-ups tend to locate close to the origins of their founders and in regionswith more investment management firms, banking establishments, and large in-stitutional money managers. New money managers show a strong local bias intheir equity holdings, three times the levels previously documented for mutualfunds. The presence of sub-advisory opportunities from institutional investorsin the vicinity is associated with the preference to base closer to home. Localbias levels between managers who relocate with their start-ups and the rest of theentrepreneurs are similar. However, preferences for stocks that were proximatewhile the entrepreneurs were in paid employment persist.

Apart from contributing to the existing research on home bias in investors'location and portfolio decisions, this paper relates closely to at least three otherstrands of the literature. First, the study partially answers Frame and White's(2004) call for researchers to examine financial innovations, including new orga-nizations. Second, the proliferation of investment management vehicles has beenthe subject of considerable interest among financial economists (e.g., Khorana(1996) and Khorana, Servaes, and Tufano (2005)). By examining the spatial dis-tribution of new investment firms, this paper extends our understanding of the ex-pansion of the investment management industry. Third, the career track records ofprofessional fund managers have increasingly caught the attention of academics(Chevalier and Ellison (1999)). Given that the existing managerial turnover liter-

'See, for an example, "Chicago blues: It's a great, livable city, but it's fading as a business andfinancial capital. What can be done?" BusinessWeek, 16 October 2000.

Parwada 247

ature generally ignores entrepreneurial investment managers, this paper, with itsfocus on self-employment as a form of managerial promotion, helps address thequestion: where do experienced fund managers go?

I organize the rest of the paper as follows. Section II describes the hypothesesthat will be tested. The compilation of the data set is summarized in Section III.Section IV examines the determinants of start-ups' location choices. Section Vanalyzes the geographic allocation of equity investments by start-ups and assessesthe relation between local portfolio bias and several location characteristics. Keypatterns of the stock selections of a subset of managers who relocate with theirstart-ups are also presented. Section VI summarizes the findings and suggestsdirections for further research in this area.

II. Development of Hypotheses

In this section, I develop two main sets of hypotheses that support the anal-ysis presented in this study. The first set concerns the characteristics of locationsthat are chosen by entrepreneurial fund managers. The second set of hypothesesis founded on the null that location factors are not related to a fund manager'spreference for locally-based stocks. I discuss each set of hypotheses in turn.

A. Determinants of Start-Ups' Location Choices

It is commonplace in studies of firm location to hypothesize that start-upsin the investment management industry seek advantages that accrue from collo-cating with other firms. One source of such advantages might be the spilloverof knowledge between incumbent firms, including buyer and supplier industries(Porter (1990)), and the start-up. On the supply of expertise, Zingales (2000) ar-gues that human capital is critical to the development of new firms. I hypothesizethat the availability of establishments in the same industry as well as related activ-ities contribute to start-ups' location decisions. The presence of related industriesreflects the potential for start-ups to tap skills outside their own industry.^ On thedemand for investment advisory services, Fischer and Harrington (1996) suggesta theory in which service-oriented firms benefit from heightened demand whenthey collocate with establishments in the same industry because clients are drawnto such areas. Thus, related industries could also capture the market potential ofthe area as conduits for funds under management from wealthy private and publicfirms, agencies, and households. Another potential source of positive returns formoney management start-ups is the presence in a location of repositories of assetsto be managed. Large investment management firms are regarded in the financialpress as a magnet for new firms. Moreover, to the extent that large institutionalinvestors reflect the maturity and innovativeness of the incumbent industry, I ex-pect their presence in a Component Economic Area (CEA) to provide a favorablebase for networking and attracting resources that foster start-ups.-'

^For an example of the exodus of expertise from the banking sector to the investment managementindustry see "Flight of SunTrust Money Management Execs Shows It's Tough for Banks to Keep TopTalent," American Banker, 160 (71), p. 1, 13 April 1995.

'For example, in comparing the contributions of large investment companies based in Bostonand Chicago to the development of money management in the cities, BusinessWeek writes: "Local

248 Journal of Financial and Quantitative Analysis

An alternative perspective to the location factors predicted above is that en-trepreneurial fund managers could actually resist clustering. One reason could bethe quest to differentiate a start-up by locating away from the competition. Seek-ing new markets could drive start-ups to base in areas with low concentrations ofcompetitors.

In the theoretical literature, high labor costs are generally regarded as dis-couraging location. However, the empirical evidence is mixed. For example,Bartik (1985) finds such costs deter investment, while Carlton (1983) reports theopposite. In the case of new investment management firms, experts admit that"people are the hands-down largest expense.'"* One prediction is that high em-ployment costs are regarded as necessary for attracting and retaining expertise.Therefore, these costs will not deter entrepreneurs. Another possibility is thatentrepreneurs know the best investment professionals are concentrated in someareas. Since their marginal product is higher, wages are higher in these areas andserve as a proxy for high quality potential employees, thus attracting start-ups.

Urbanization is also regarded as a strong allure to start-ups in the literature, Iexpect the extent of a region's urban infiuence to affect location decisions. For in-stance, financial centers may induce fund managers to be overconfident in their in-vestment decisions (Christoffersen and Sarkissian (2004)). Overconfidence abouta firm's prospects could affect managers' location decisions (Almazan, de Motta,and Titman (2003)), Extending this logic to investment management start-upsraises the possibility that this phenomenon and other unobservable factors uniqueto financial centers could infiuence start-ups' location choices. Managers mayalso be guided by property cost information in finding a location, resulting infewer start-ups being based in financial centers.

The assumption implied in these predictions is that the location factors alsomotivate fund managers' decisions about whether to stay in the area where theyheld their last job or move away. Alternatively, fund managers may opt not to re-locate when they start their own firms for personal social and economic reasons.Such infiuences could be thought of as characterizing the constrained nature ofa manager's location choice.^ For example, a major consideration in whether tochange locations is likely to be the localized nature of a portfolio manager's rep-utation. Personal familiarity appears to be an important determinant of relationsbetween fund managers and their clients. From anecdotal evidence, while perfor-mance records are useful, fund managers enjoy a significant measure of trust fromtheir local peers and clients based on past relations. These networks may becomeimportant considerations in location decisions.* Other considerations include per-

fund managers blame the disappointing comparisons on lost opportunities. At Stein Roe, the ownersearly on balked at hefty investments in marketing and infrastructure, argues Marshall B, Front, itsex-president, 'Stein Roe wasn't much different in size from Fidelity and Vanguard (in the 1970s),' hesays, 'Those organizations were more forward thinking. They saw the potential of mutual funds,'"BusinessWeek, 16 October 2000,

•'See "Spinning Off a Firm on Their Own," Investment Management Weekly, 11 May 1998,' i thank the referee for suggesting this interpretation,*For example in the article, "Loyalty to Detroit Area Pays Off for Investment Entrepreneur," Amer-

ican Banker, 152 (166), p, 6, 25 August 1987, Lee Munder is reported as saying: "Professionally,Detroit has been a wonderful community to me. Clients here tend to be very loyal. There is a mutualrespect among competitors, and a bond exists within the financial community that is indigenous onlyto the Midwest,"

Parwada 249

sonal ties to family and friends, schooling arrangements for children, and the costof relocating.

To take these issues into account, I hypothesize, first, that social and lifestyle-related factors will affect the decision of where a new fund firm is located. Sec-ond, I expect that there will remain influences that cannot be disentangled fromthe general advantages of a manager's familiarity with a certain location. For ex-ample, there could be costs involved in relocating that are large enough to bias thelocation decision toward the home base. Such costs may not be fully reflected invariables relating to individual managers' lifestyles and social preferences. Third,factors that attract fund managers could present stronger influences on managerswho move from their home region (movers) than those who remain in their homebase (non-movers) (Figueiredo et al. (2002)). These considerations suggest theneed to control specifically for the home base when considering determinants oflocation choice.

B. The Extent and Determinants of Local Bias in Start-Ups' Portfolios

The first question of interest with regard to the geographic coverage of thestocks held by fund start-ups is whether local bias in entrepreneurial fund man-agers' portfolios differs from levels documented for mutual funds in studies suchas Coval and Moskowitz (1999), (2001), Hong et al. (2005), and Ivkovid andWeisbenner (2005). If entrepreneurs remain within their regions of origin so asto continue to exploit local networks, one might expect them to exhibit at least asmuch home bias in their stock preferences as their previous employers. Similareffects could emerge if the new firms replicate the (home-biased) strategies of theformer employers of their founders. Finding higher levels of home bias for thestart-ups could signal that new firms, facing restricted access to resources, gener-ally avoid research-intensive distant stocks. In such a case, they would rely moreheavily on local information as a low cost strategy for competing with establishedinstitutions. This argument conforms to the "informational asymmetry" hypothe-sis predicting that local investors tilt their portfolios to nearby stocks because theyhave superior information on such securities (see Coval and Moskowitz (2001)).

An alternative argument can be mounted against the local bias hypothesis.Investment company start-ups are predominantly institutional account managers.Thus, their local bias could be lower than the levels documented for mutual fundssince institutional clients are less prone to biases in their portfolio decisions thanindividual investors (in mutual funds) (Shapira and Venezia (2001)).

The geographic spread of a start-up's operations, proxied for by stock hold-ings, may contain information on location choices. Next, I turn to the questionof whether factors affecting location decisions manifest themselves in start-ups'portfolio decisions as well. I start with the null hypothesis that they do not.However, Goetzmann, Massa, and Simonov (2004) suggest that the knowledgespillovers accompanying the clustering of firms induce investors to choose famil-iar assets. In managing their portfolios, entrepreneurial fund managers primarilyrequire research expertise and capital. Therefore, the supply of related establish-ments and institutional investors may become important influences on the deci-sion to bias portfolios close to or further away from the home base of the start-up.

250 Journal of Financial and Quantitative Anaiysis

I also expect that the social and economic prospects of an area will be impor-tant indicators of the profit potential of companies located in close proximity to astart-up and will be positively related to the extent of local bias in their portfolios.

If employment costs are an indicator of the quality of expertise available inan area, they could predict the extent of home bias in managers located there. Forexample, finding a negative relation between labor costs and local bias in an area'sfund managers would imply that better remunerated managers are less prone torelying on home bias in their portfolio strategies.

Fund managers located in financial centers are more likely to select morestocks in their immediate vicinity than those located in smaller cities (Coval andMoskowitz (2001)). Moreover, Goetzmann et al. (2004) show that urban investorsare more locally biased than rural-based investors. Following these cirguments, Ialso hypothesize that location factors defining the extent of urbanization in an areawill be pertinent to the geographic spread of start-ups' portfolios.

In the context of start-ups in the investment management industry, it seemsplausible that if local bias relates to area characteristics the link should be moresalient for managers who retain their home base than for movers. To the extentthat local bias is a result of the exploitation of local networks and resources, thedistinction between movers and non-movers cannot be ignored. My hypothesesconcerning the influence of location characteristics on local bias therefore lead toa requirement for the home base of the entrepreneur to be controlled for.

Finally, that some entrepreneurial fund managers relocate to distant areasupon forming their own firms while others stay in their home base raises two ques-tions regarding the geographic spread of start-ups' portfolios. First, do movers re-locate to areas with high or low competition in terms of the number of fund man-agers relative to the stocks located in such areas? Fischer and Harrington's (1996)argument that consumer demand and collocation with related industry firms arelinked seems sufficient to yield the hypothesis that movers in this study will basein areas with high concentrations of fund managers relative to stocks. An alter-native to this hypothesis is that movers choose areas where there are relativelymany stocks and few fund managers (e.g., Bellevue, WA and Menlo Park, CA).^Second, do movers show a tendency to hold on to stocks that were proximate totheir former home bases? Answering this question is a direct test of whether localnetwork effects persist after a manager relocates to start a new firm and consti-tutes a further robustness check on the role of geography in the genesis of homebias. I expect that at least earlier in their firm's life cycle, fund managers willbe affected by a "look back" bias toward stocks they were previously familiarwith. Furthermore, as suggested by the home bias literature, such stocks wouldbe located close to their former home bases. In the meantime, if there are indeedadvantages to biasing stock selections locally, it is reasonable to expect a bias infavor of the new locations to start developing as well.

' i thank the referee for suggesting these examples.

Parwada 251

III. Constructing the Data Set

A. Start-Up and Parent Information

Critically, the study requires tracing the locations of fund managers' last em-ployers and of their own firms. The data collection exercise is complicated by thefact that commonly used mutual fund databases such as Morningstar and CRSPdo not contain information on reasons for or dates of manager turnover at estab-lished fund companies as well as their background career histories. Hence, thesedatabases cannot be utilized to identify a sample for this study. I use a uniquehand-collected data set culled from a combination of public domain sources andelectronic databases. To generate the initial list of fund managers that leave estab-lished investment companies specifically to set up their own money managementfirms, I search the Lexis-Nexis, Factiva, and ABI-Inform databases for relevantnews articles. I supplement these sources with searches of ADV Forms lodged byfund advisors when they register for licensing through the SEC's Investment Ad-visor Public Disclosure system. I also identify start-ups and their founders in fundmanager directories and Web sites of mutual fund companies. The resulting dataset is a rich history of new mutual fund companies at their inception detailing thefund managers, parent fund families, departure dates, and identities of start-ups. Ialso collect information on the locations of parent and start-up firms.

Several advantages can be ascribed to the data set construction process. First,I sample start-up companies at birth, hence limiting potential survivorship bias.Linked to this advantage, observing start-ups at birth largely resolves concernsabout subsequent name changes as both parents and new firms are identified withthe names that exist then. Second, the data set covers periods of varying levelsof popularity of fund types and styles, meaning that my results will likely not besolely driven by transient trends.

One natural concern about using news articles as the primary source of infor-mation on entrepreneurial activities is that the sample is biased toward successfulfund managers likely to receive media coverage. This concern is resolved, at leastin part, by the fact that my search covers both high profile national news outletssuch as The Wall Street Journal, presumably prone to following high fiying port-folio managers, and smaller regional publications that are likely to report on fundmanagers of smaller stature.

For purposes of this paper, I define entrepreneurs as the founders and initialexecutive officers in firms formed by investment professionals leaving establishedfund families. The original sample of fund managers comprises over 450 fundmanagers spanning the period 1980 to 2003. However, matching this data set withlocation characteristics as described below reduces the number to 358 executivesat 207 firms founded from 1988 to 2003.

B. Locality Information

Tests on the determinants of location choice at the firm level are based onthe CEA as defined by the Bureau of Economic Analysis (BEA). Throughoutthe rest of the paper, I use the terms CEAs, areas, and regions interchangeably.The 348 CEAs covering the U.S. land mass are collections of counties centered.

252 Journal of Financial and Quantitative Analysis

in most cases, on metropolitan statistical areas (see Syverson (2004) for a de-tailed explanation). While a drawback of using this unit of analysis may be thatit imposes arbitrary boundaries on the location characteristics considered by en-trepreneurs, the BEA selects counties for inclusion in a given CEA on the basis oftheir metropolitan statistical area status, worker commuting patterns, and news-paper circulation patterns, subject to the condition that CEAs must contain onlycontiguous counties. Thus, the selection process ensures that counties in a givenCEA are economically related (Johnson (1995)). Therefore, I argue that the CEAis a suitable compromise to using state boundaries that could be considered toobroad or county borders that could be too localized. More importantly, usingCEAs gives a geographic unit with boundaries that are consistent throughout thispaper's sample period.^

I link all CEAs to the BEA's County Business Patterns Database based onthe Business Information Tracking System (BITS) database. Introduced to the fi-nance literature by Villalonga (2004), BITS is a longitudinal database that tracksindividual firms and their employment growth over time and constitutes the eco-nomic micro data aggregated at the county level in the County Business PatternsDatabase. Therefore, for each CEA I am able to track mid-March employment,payroll, employee numbers, and establishment count data for the period 1988 to2002, the last year in the database. Although the basic unit of the BITS datais a business establishment (location or plant), at the County Business Patternsdatabase level, the aggregation process scrambles individual firm identities. Inthe event of sparse data at the county level, making it possible to link the datato individual establishments, the BEA camouflages firm identities by replacingnumbers with alphabetic flags representing employment size class (at ranges of1-4, 5-9, and so forth) for data withheld to avoid disclosure. In such cases, thenumber of employees and payroll data are replaced by zeroes. To estimate em-ployment size for these firms, I follow Syverson (2004) and take the midpoint ofthe size class given in the database. In the case of withheld payroll data, I turnto the comprehensive data on number of establishments and compute the averagenumber of employees per establishment. I am then able to obtain an estimate ofthe annual payroll per employee by dividing the average payroll amount by thenumber of employees per establishment.^

I collect county level information from the County Business Patterns databaseand aggregate the data at the CEA level by industry as follows: investment man-agement (SIC/NAICS codes 6700 (before 1998) and 524 (from 1998 onward)),securities (6200, 523), banking (6000/6100, 522), and insurance (6300, 524).'° I

*1 utilize the 1995 BEA economic area boundaries (see Johnson (1995) for details). For the pre-1995 period, I simply treat the current boundaries as applying them to avoid problems of data discon-tinuity due to shifting CEA boundaries.

'Although it might be preferable to base payroll estimates on full-time equivalent employees, theBEA lumps full- and part-time workers together. The result is that the average employment costs forindustries and CEAs with high numbers of part-time people may be lower than those with a higherpercentage of full-time employees. However, for purposes of this study the important issue is todifferentiate regions on the basis of employment costs and not on determining the absolute levels ofsuch expenses.

'"Beginning in 1998, the data are tabulated by industry as defined in the North American IndustryClassification System: United States, 1997 (NAICS). Data for 1997 and earlier years are based onthe Standard Industrial Classification (SIC) System. 1 use SIC to refer to both classification systems

Parwada 253

also obtain data on the dominant investment management firms in the U.S, fromthe Institutional Investor America's Top 300 Money Managers Surveys of 1988,1992, 1996, and 2001, I assign assets reported in millions of dollars to relevantCEAs according to the location of the institutions as disclosed in the surveys.

For further social and economic CEA characteristics, I use the U.S. Depart-ment of Agriculture (USDA) County Typology Codes," In the data, reflectingarea characteristics, Metro (or Non-Metro), Housing Stress, Low Education, LowEmployment, Persistent Poverty, Low Employment, Recreation, and Retirementdominant counties are marked by 1,0 dummy variables. The Urban Influencecode ranges from 1 (highest) to 12 (lowest). I aggregate the county typology datafor each region by taking an average of each indicator for all counties constitutingthe CEA.

I describe the spatial distribution of start-ups in Table 1. Panel A shows thatthe top 10 localities of choice (hosting four or more new firms) are dominatedby financial centers with New York, Boston, San Francisco, and Chicago occu-pying the first four positions. Further, as reported in Panel B, start-up numberscoalesce in 26 states led by California (36), New York (32), Massachusetts (27),and Pennsylvania (16).

Table 2 compares CEAs that are chosen as bases by new firms to those thatare never selected. I take the average for each CEA that ever hosted a new firmduring the sample period: i.e., those that were chosen as locations by at least onestart-up. I contrast these CEAs to those that never host a start-up over the timeperiod. The tabulation shows that 140 out of 348 CEAs are chosen as locationsby entrepreneurial fund managers and selected regions tend to be considerablylarger than non-hosts. For instance, the median non-hosting CEA is about 5%the size of the median hosting CEA in terms of investment industry employmentnumbers and 10% on the basis of establishment counts. Similarly, CEAs that arechosen by entrepreneurs are home to significantly larger banking and insuranceundertakings. However, investment industry employment costs in the locationsthat entrepreneurs choose are on average nineteen-fold those of non-host CEAs.The median start-up-hosting CEA is also the location of institutions controllingover $20 billion in assets among the top 300 investors in the country, compared toa median of zero for non-host CEAs.

I report the spatial distribution of entrepreneurs relative to their origins inPanel A of Table 3. The strong preference for investment firm founders to locatein close proximity to their origins is very clear. Entrepreneurs that stay withinthe CEA where their last employer is located comprise 80% of the sample. Evenwhen I consider locations at the county level, 68% of the entrepreneurs choosethe county hosting their last place of employment. To give another perspectiveof the geographic distribution of entrepreneurial fund managers relative to theirorigins, I use distance in kilometers as an alternative measure. To each location,usually identified in source documents by city and then state, I assign latitude andlongitude coordinates obtained from the Geographic Names Information System

for simplicity. My reference to the word industry is also restricted to the financial services sectors ofbanking, investments, securities, and insurance,

' ' See http://www,ers,usda,gov/Data/TypologyCodes/,

254 Journal of Financial and Quantitative Analysis

TABLE 1

Summary Information on the Location of Start-Ups

Table 1 summarizes the locations of 207 firms started by former investment managers during the period 1988-2003. Thesampie is identified from searches of public domain sources and electronic databases. The table reports start-up countsby location and state in Panel A and by state only in Panel B.

Panel A. Number of Start-Ups by Location

Location

New YorkBostonSan FranciscoChicagoMilwaukeeLos AngelesStamfordMinneapolisSah DiegoSan fviateoBaitimoreCambridgeDenverGreenwichMalvernNewport BeachArlingtonBelievueBerwynChattanoogaFort LauderdaieGreenwood ViliageHoustonJacksonviiie BeachLouisvilleOrlandoPhiiadelphiaSan Antonio .WeilesleyAiexandriaAhdoverBala Cynwyd

State

NYMACAILWlCACTMNCACAMDMACOCTPACAVAWAPATNFLCOTXFLKYFLPATXMAVAMAPA

Start-Ups

282011107554443333332222222222222111

Panel B. Number of Start-Ups by State

State

CaliforniaColoradoConnecticutWashington DCDenverFloridaIllinoisKansasKentucky

Code

CACOCTDCDEFLILKSKY

Start-Ups

366

11 '118

1112

Location

BeachwoodBethesdaBioomfield HiilsBiue BeilBoca RatonBrookfieidChadds FordCharlotteCincinnatiCiaytonCieveiandConshohockenDallasDaytonDel MarDurhamEnglewoodGlendaleGuiifordIndependenceIrvineKansas CityKansas CityKing of PrussiaLisleLos AltosMenashaMenio ParkMequonMinnetonkaNashviilePasadena

State

MassachusettsMarylandMichiganMinnesotaMissouriNorth CaroiihaNew JerseyNevadaNew York

State

OHMDMlPAFLWlPANCOHMOOHPATXOHCANCCOCACTOHCAKSMOPAILCAWlCAWlMNTNCA

Code

MAMDMlMNMONCNJNVNY

Start-Ups

1111111T111111111111111111111111

Start-Ups

275354341

32

Location

PittsburghPrincetonPurchaseRichmondRochesterRocklandRockvilleRosevilleSaint ChariesSaint LouisSan RafaelSanta AnaSanta BarbaraScituateShrewsburySouthportSparksSummitTampaTroyTuaiatinWashingtonWayneWestportWexfcrdWhite PlainsWinston-SalemWoodoliff LakeYardieyYonkers

State

OhioOregonPennsylvaniaTennesseeTexasVirginiaWashingtonWisconsin

State

PANJNYVANYDEMDMlMOMOCACACAMANJCTNVNJFLMlORDCPACTPANYNCNJPANY

Code

OHORPATNTXVAWAWl

Start-Ups

111111111111111111111111111111

Start-Ups

51

163542

10

(GNIS).'^ I calculate distance between two places using their respective coordi-nates in spherical geometry as follows:

(1) d{A,B) = 6370.997 X {arccos [sin (latitudeA) X sin {latitudeB)

X cos {latitudeA) X COS {latitudeB)

X cos {\longitudeA —

'^Administered by the U.S. Department of the Interior/U.S. Geological Survey and available onlineat http://geonames.usgs.gov/.

Parwada 255

TABLE 2

Descriptive Statistics: Characteristics of CEAs Hosting Start-Ups vs. Non-Host CEAs

Table 2 compares characteristics of Component Economic Areas (CEAs) selected as locations by start-ups (the first threecolumns) to the characteristics of CEAs that are never chosen during the sample period (the next three columns). Thesample comprises firms started by former investment managers who depart from their empioyers from 1988-2003. Thesample is identified from searches of public domain sources and electronic databases. Data on CEAs are obtainedfrom the Bureau of Economic Analysis. Data aggregation under economic sectors (bank, investment, and insurance) isbased on the County Business Patterns database. Employment costs are calculated as annuai average payroil costs peremployee (part-time and fuii-time). Top 300 institutional assets and companies are obtained from fhe Institutional InvestorAmerica's Top 300 Money Managers surveys.

CEAs Chosen CEA Not Chosenfor Location for Location

Number of start-upsBank employmentInvestment employmentInsurance employmentBank establishmentsInvestment establishmentsInsurance establishmentsBank payroil costsInvestment payroll costsInsurance payroll costsTop 300 institutional assetsTop 300 institutional companies

Mean

130,17012,41520,555

1,476802888

54,72967,94438,656

152,2225.03

Median

123,3385,737

15,2051,235

496532

37,60226,31518,60920,744

2.00

Obs.

140140140140140140140140140140140140

Mean

4,154765

2.779285

88163

7,6014,0285,4724,165

0.18

Median

2,2782998811824980

4,1011,3682,038

00

Obs.

208208208208208208208208208208208208

where latitude and longitude are measured in radians.''' The constant, 6370.997,is the earth's radius in kilometers and converts the distance into units of one kilo-meter.

Panel B of Table 3 gives the cumulative distribution of distances betweenfirm founders and their origins. Of note is the large proportion, 56%, of en-trepreneurs locating within a kilometer of their last employers. Seventy-five per-cent locate within 25 kilometers of their origins. These figures closely corrobo-rate the impression of "home-bias" in location decisions given by the CEA-basedanalysis.

C. Portfolio Holdings

It is common for fund managers that establish their own firms to start as insti-tutional account managers. Money management firms that accumulate significantassets (exceeding $100 million) have to submit 13f Holdings returns to the SEC.I use the SDA/Spectmm database provided by Thomson Financial to access theportfolio holdings of 152 start-up investment management firms from my originalsample in the period 1984 to 2003. Of these companies, 141 are in the 1988-2003 sample used up to this stage. I concentrate on holdings at the end of thefirst year of a fund company's appearance in the database. In this way, I achievea specific cut-off point that is not reliant on the varying dates at which the firmsstart reporting holdings data. It is also necessary to restrict myself to holdings thatare as close to the commencement of the firm's operations as possible, in keepingwith the concern of this paper with decisions taken by founders of start-ups at thebeginning of the firms' existence.

"See Ivkovid and Weisbenner (2005) for a similar application.

256 Journal of Financial and Quantitative Analysis

TABLE 3

Distribution of Entrepreneurs' Location Choices Relative to Origins

Table 3 provides information on the distribution of founders of investment companies. The sample oomprises investmentmanagers who depart from their employers to form their own firms from 1988-2003. identified from searohes of publicdomain sources and eieclronic databases. Each entrepreneur's origin is set as her place of iasf employment. PanelA reports summary information on founders that do not relocate and those that change locations. Panel B gives thecumulative distribution of the distanoe between founders and their origins. The distance between an entrepreneur and herorigins is calculated using spherical geometry from the coordinates of fhe places.

Panel A. Distribution of Start-Ups'Locations Relative to Origins

Founders in original looafiohFounders out of original location

Number

CEA-Level County-Level CEA-Level

281 239 80%71 113 20%

Panei B. Cumuiative Distribution of Distance to Origins among investment Management Start-Ups

Distance(kilometers)

15

10152025304050

100500

1,0002.0004.200

Number ofEntrepreneurs

1982172342492542652742893033143263413 4 5 . • •

352

Percent

County-Level

68%32%

Proportion

56%62%66%71%'72%75%78%82%86%89%93%97%98%

100%

IV. Location Determinants

A. Basic Specification and Definition of Variables

I start my analysis of location determinants with the traditional conditionallogit framework based on McFadden's (1974) choice model and utilized in theindustrial location literature dating back to Carlton (1983). The basic approachconsists of treating the location decision problem as one of random profit max-imization. Given a set of mutually exclusive CEAs, the entrepreneur weighs inall the location characteristics of the available spatial choice set and selects theone that will potentially give her the highest profit. Following McFadden (1974),if I assume the error terms are distributed independently and follow an extremetype value I distribution, I end up with the logistic formulation. In this paper, afirm's founders choose from the CEA in the sample. The dependent variable istherefore denoted 1 for each CEA chosen by a new investment advisory firm in agiven year, and 0 otherwise. I discuss the independent variables in detail below.

Among the independent variables, the factors relevant to this analysis areexternal economies assumed to lead to the clustering of investment managementfirms. The first set could be interpreted as being relevant to location decisionsat three levels: providing markets, being a source of specialized labor, and sim-ply fostering new ideas and innovations that spill over from collocated firms. Themarkets for an investment management firm are defined by the source of funds un-der management, the commonly used basis for compensating portfolio managers.

Parwada 257

I incorporate the number of establishments in the investment sector (investmentsand securities SICs) located in each CEA. I take the natural logarithm of the num-ber of establishments plus 1, to avoid disregarding CEAs with no establishmentsin this industry. I also include similar metrics for the banking and insurance SICs.

For each CEA, I record the natural logarithm of the total assets managed byfirms in the Institutional Investor America's Top 300 Money Managers, sampledin the years 1988, 1991, 1996, and 2001. The market potential arguments forincluding this variable are similar to the ones for industrial sector establishmentdata. In short, a fund manager's success as an entrepreneur is often measuredin terms of how quickly she secures assets under management. According to myhypotheses regarding locating start-ups close to resources, managers that perceivean advantage in locating close to large institutional investors will give this factorsubstantial weighting in their location decisions. I also include an indicator of thelevel of employee costs in different locations denoted by the natural logarithm ofthe average payroll per employee as described in Section III.

The model contains dummy variables representing Boston, Chicago, LosAngeles, New York, Philadelphia, and San Francisco. These cities have beenidentified as being dominant fund centers in studies such as Christoffersen andSarkissian (2004) and Hong et al. (2005).

County typology factors in the model pick up two effects motivated by myhypotheses. First, from the perspective of an entrepreneurial fund manager, fac-tors such as Low Employment and Low Employment complement indicators ofthe profit potential of the area. Second, the variables attempt to capture the per-sonal economic and social preferences of fund managers who make the choiceto become entrepreneurs. The specification includes the variables Metro, UrbanInfluence, Financial Center, Housing Stress, Low Education, Low Employment,Low Employment, Recreation, and Retirement as described in Section III. .

Econometric Approach. Some econometric concerns arise from the natureof my data. The dependent variable is a binary indicator with several subjects(CEAs) repeated in the event of more than one start-up choosing the same loca-tion. A significant portion of the sample also involves CEAs that do not attractstart-ups for extended periods. Therefore, in CEA-based tests I augment standarddiscrete choice models with logistic models that account for the clustered natureof the data by incorporating the Generalized Estimation Equation (GEE) tech-nique. GEE is an extension of generalized linear models applied to longitudinaldata (Liang and Zeger (1986)). GEE corrects for clustering among correlated vari-ables by increasing the correlation among observations within a cluster relative tothe correlation between clusters. The results are qualitatively consistent betweenthe basic logistic regression models and their GEE versions. The effect of GEEis to lower the size and significance of the coefficients marginally. However, inno case is there a sign change between logistic and GEE models, reinforcing therationale for using GEE. Since the procedure yields results that are more conser-vative than those from models that do not consider the natural autocorrelation ofdata similar to. those applied in this paper, I only report the GEE estimates.

258 Journal of Financial and Quantitative Anaiysis

B. Empirical Results

Basic Specification. The findings of the empirical analysis in this section arepresented in Table 4.1 split the analysis of the basic specification into two modelsreported in the first and second columns of Table 4 and labeled Models 1 and 2,respectively. The models examine the effects of Investment Establishments andTop Institutional Assets separately. Establishments could offer more informationabout the availability of expertise in an area while institutional assets could relatemore closely to a start-up's prospects of raising funds to manage. The findingsregarding location determinants are largely consistent with the hypotheses moti-vated in Section II.A.

Two results stand out in the basic models. These are the tendency of start-ups to collocate with incumbent firms and the effects of social and economicarea attributes. In Model 1, the coefficients on the establishment variables for thebanking and investment industries are positive and statistically significant. As pre-dicted, investment management start-ups are attracted to areas with a significantpresence of other firms in the same industry, and are likely to take advantage oflocal industry networks. They also collocate with bank undertakings. Bank trustdepartments house well-trained investment professionals with a history of cross-ing over to the investment industry. There is also weak evidence that InsuranceEstablishments discourage collocation with investment management start-ups. InModel 2 of Table 4, the coefficient on the variable measuring institutional assetsunder management. Top Institutional Assets, is positive and highly statisticallysignificant. This finding suggests start-ups do chase this rich source of funds un-der management in their location decisions. Regarding employment costs, ratherthan being deterred start-ups select regions with high payroll costs. This resultcould be a reflection of a realization by founders that high employment costs areunavoidable when trying to enter this industry.

The second outstanding feature of the results is that, as expected, low edu-cation and low employment areas are shunned by start-ups. A surprising resultis the neutrality of the Financial Center dummy. One would intuitively expectnew investment advisory firms to locate in financial centers, not least to chaseassets under management. Finding the opposite result could be related to increas-ing preferences in services industries for suburban office locations at the expenseof financial centers. Although the coefficient on Metro suggests metropolitan ar-eas are favored for start-up locations, the variable has a low level of statisticalsignificance.

The Role of the Home Base. The factors considered in the basic specifica-tion could be correlated with a fund manager's consideration of the profitabilityof an area. However, as argued in Section II.A, there may be information in theinfluence of where the founders of the new firms originate. While the variablesrepresenting social and economic characteristics could account for personal fac-tors that are important to the fund manager, they might not be sufficient to accountfor the full extent of the influence of her origins. In Models 3 and 4 of Table 4,following Figueiredo et al. (2002), I include the variable Home Base to representthe CEA of origin of each firm. Because multiple founders can emanate from dif-ferent areas, the Home Base dummy equals 1 where more than half the founders

Parwa(da 259

TABLE 4

Factors Affecting the Location Decisions of Investment Management Start-Ups

Table 4 reports Generalized Estimation Equation regressions of investment management start-ups' choice of ComponentEconomic Area (CEA) in tiie period 1988-2003. Estabiisfiment counts and employment costs (average annuai payroliper employee) are from the Bureau of Economic Analysis. Top Institutional Assets field in each CEA are from InstitutionalInvestor America's Top 300 f^ioney f^anagers surveys. Financial Center is an indicator of CEAs incorporating Boston.Chicago, Los Angeles, New York, Philadelphia, or San Francisco. Metro, Housing Stress, Low Education, Low Employ-ment. Low Employment, Recreation, and Retirement county indicators and the Urban Influence code (ranging from 1(highest) to 12 (lowest)) are obtained from the U.S. Department of Agriculture. Home Base is a 1,0 dummy variable indi-cating a majority of a start-up's founders originate from that CEA. Standard errors are in parentheses.'",", and ' denotestatistical significance at the 1%, 5%, and 10% levels, respectively

fviodel 1 Model 2 Model 3 Model 4

Code

Bank Establishments

Investment Establishments

Insurance Establishnwnts

Top tnstitutionai Assets

Bank Payroll Costs

Investment Payroll Costs

Insurance Payroll Costs

Metro

Urban Influence

Financial Center

Housing Stress

Low Education

Low Employment

Low Employment

Recreation

Retirement

Home Base

Log Lii<eiihood

15.5004*"(1.7272)

1.4491***(0.4914)

0.5190**(0.2533)

-1.1646*(0.6746)

-0.4504(0.3306)

0.1853(0.1803)

0.8726(0.5512)

1.7211*(0.9861)

0.1072(0.1625)

-0.1826(0.3498)

-0.3554(0.2607)

-0.8680**(0.3766)

-1.9505**(0.9020)

-0.3064(1.0731)

1.1274(0.8618)

-0.6283(0.6298)

-471.34

14.6402"*(1.8961)

1.5368*"(0.4788)

-1.0714(0.6755)

0.1242***(0.0438)

-0.4402(0.3103)

0,2873**(0.1468)

0.7694(0.5446)

1.4113(0.9503)

0.0967(0.1621)

-0.0453(0.2814)

-0.0718(0.231)

-0.9309***(0.3227)

-1.1813*(0.6341)

-0.1592(0.8702)

1.0763(0.8306)

-0.5969(0.7197)

-464.23

9.2510***(1.4335)

1.0011***(0.3537)

-0.0724(0.2062)

-0.7697*(0,4232)

-0.2754(0.2029)

0,0848(0.1022)

0.6666**(0.3248)

-0.2447(0.6831)

-0.0541(0.1126)

-0.6864***(0.2423)

0.3443*(0.1936)

-0.3789(0.2730)

-2.0338***(0.6601)

-0.2316(0.8383)

-0,7317(0,7240)

0,3789(0,4346)

2,7552***(0,4783)

368,67

8.4700***(1.2618)

0.8843**(0.3286)

-0.7893*(0.4537)

0.0608(0.0401)

-0.3428*(0.1828)

0.0457(0.0783)

0.6858*(0.3512)

-0.1841(0.718)

-0.0454(0.1214)

-0.5451**(0.2225)

0.3818**(0,1684)

-0,3670(0,2774)

-2 ,1519*"(0,5979)

-0.1202(0.8017)

-0.7823(0.7674)

0.4330(0.4455)

2.6994***(0.4790)

365.31

of a start-up come from the same locality as that of the firm, and 0 for all others.The fit of my estimates improves significantly as shown by the substantial changein the log-likelihood value. The coefficient estimate of the Home Base dummyvariable is sizeable and statistically significant and its positive sign suggests anoverwhelming influence of the entrepreneur's origins on where a start-up locates.

Notably, a number of coefficient estimates that were significant in Models1 and 2 of Table 4 maintain their signs and magnitude. Some are reduced inmagnitude, while other variables pick up more significance. After incorporatingentrepreneurs' prior locality indicator. Bank Establishments continue to exert a

260 Journal of Financial and Quantitative Analysis

positive infiuence on an area's likelihood of being chosen. Insurance Establish-ments deter collocation and low employment areas are also avoided. However, ap-parently absorbed by the Home Base dummy variable, the influence of InvestmentEstablishments and Top Institutional Assets is still positive but much diminishedas shown by the loss of statistical significance on the coefficient estimates. Giventhat the majority of entrepreneurs in my sample do not relocate, this result maybe interpreted as indicating that the benefits to an entrepreneur of collocating withincumbent firms and repositories of core assets to be managed are imbedded in theadvantages associated with staying in the locality of origin. Similarly, Low Edu-cation is likely a factor closely associated with the choice of whether to remain inan area and is thus also reduced in its influence by the Home Base dummy. It isalso possible that factors unrelated to the decision to stay in the same location arethe ones that stand out separately even in the presence of prior locality consider-ations. For example, in both Models 3 and 4 of Table 4, high payroll costs in theinsurance industry are now positively related to the probability of a region beingchosen while financial centers are a deterrent.

V. Location Choice and Local Bias in Portfolio Preferences

Having analyzed the firm location decisions of entrepreneurial fund man-agers, I also consider relations between their new firm's location and stock hold-ing preferences. As noted in the Introduction, strong evidence points to a homebias in the portfolio decisions of professional money managers and individual in-vestors. The bias is economically significant in terms of magnitude as well asreturns attributed to the practice, suggesting the validity of the argument that in-vestors enjoy an informational advantage when they trade in nearby stocks. Inthis section, I estimate the extent of home bias in a sample of funds managementstart-ups and relate the measures to several location characteristics.

A. Home Bias in Start-Ups' Equity Holdings

First, I compare the extent of home bias in the portfolios of entrepreneurialfund managers with Coval and Moskowitz's (2001) findings for mutual funds.I link the location of start-up firms to the location data (county and state) onheadquarters of stocks held by the firms obtained from the Center for Researchin Security Prices/Compustat merged database (CRSP/Compustat). To determinethe distance between the fund and each stock held at the end of its first year ofreporting, I employ GNIS coordinate information as described in Section III.B. Ifollow Coval and Moskowitz (2001) and define local bias in the equity holdingsof fundy at the end of year r as follows:

(2) LocalJBiaSj^, = Local-Own-PortfoliOj, — LocalJidarket-Portfolioj „

where LocaLOwti-Portfolioj^ is the fraction of the market value of a fund's to-tal holdings comprising stocks located within 100 kilometers of the firm andLocalMarketJ'ortfoUOj, is the proportion of all CRSP/Compustat listed stocks(or the market) based within 100 kilometers of the start-up. For comparison.

Parwada 261

departing slightly from the Coval and Moskowitz home bias measure, I also com-pute an alternative metric where local bias is conditioned only on the simple ruleof whether the firm holds a particular stock, rather than the value of the full posi-tion in that stock.

In Table 5, I report results of tests of equality between local bias in start-ups versus the market. On the basis of value of assets held, start-ups bias theirholdings 2.5 percentage points in excess of the 6.8% allocation to local stocksthat would have prevailed had they invested in the market portfolio. This figure isabout three times the 0.8 percentage point bias over and above 6.16% in the mar-ket documented for mutual funds by Coval and Moskowitz (2001). Redefininglocally held based on whether a stock is held by a fund, the bias is even strongerat 3.5 percentage points, suggesting that on a stock-by-stock basis start-ups areprone to higher levels of local bias than would be suggested by the value of bold-ings. These findings conform with the hypothesis that at the start-up stage, fundmanagers utilize local information in building their portfolios to an extent thattranslates into a local bias that is significantly higher than that previously reportedfor mutual fund managers. This phenomenon is observed in spite of the fact tbatthe start-ups in this study represent the interests of institutional investors whowould be expected to be less locally biased than retail clients in mutual funds.To obtain an indication of the economic significance of tbe local bias shown bystart-ups in this paper, consider the performance of locally-based stocks in Covaland Moskowitz (2001). While Coval and Moskowitz find a "modest" net localbias of 0.8% on average, local stocks held by mutual funds in their sample earna premium over the risk-free rate of 8.7% per year, 2.67 percentage points abovetheir premium on distant equities. On a risk-adjusted basis, the returns amount to184 basis points per year.

TABLE 5

Composition ot Start-Ups' Portfolios by Locality

Table 5 reports average local and non.local portfolio shares for 152 start.up investment management firms (or 267founders). For each start-up, portfoiio hoidings are extracted in the last quarter of the first year of statutory reportingof hoidings as recorded in the SDA/Spectrum database in the period 1984 to 2003. Bias.Own.Portfolio denotes the frac-tion of a start-up's total holdings comprising stocl(s located within 100 kiiometers of the firm and Bias-Market.Por{foiio isthe proportion of ali CRSP/Compustat iisted stoci<s (or the mari<et) that are based within 100 kiiometers of the start-up.Hoidings are based on doiiar value of assets or identified on a simpie ruie of whether a stock is heid by the start-up. Locaibias is the differenoe between the start-ups' and market fractions of stooks heid iocally. * " and * indicate non-equalitybetween start-ups' and market portfolios held locaily at the 1% and 10% significance ieveis.

Bias.OwnJ'ortfoiio BiasMarket-Portfoiio Difference

Value of assets residing < 100 kiiometers 9.30% 6.80% 2.50%*Number of stocks residing < 100 kiiometers 9.10% 5.40% 3.70%"*

B. Location Characteristics and Home Bias in Portfolio Holdings

The second objective of this section is to explain home bias in tbe equitypreferences of entrepreneurs in relation to the new firms' location preferences.To test the hypotheses motivated in relation to tbe determinants of local bias instart-ups' portfolios, I estimate the regression.

262 Journal of Financial and Quantitative Analysis

(3)

+ j^Insurance Establishmentsj^, + '^^Bank Payroll CostSj,

Payroll CostSj, + -y^Insurance Payroll CostSj,

InstitutionalAssetSj, + ̂ gMetroj^, + jgUrban Influence^,

+ 'jioFinancial Centerj^, + 711 Low Employmentj,

on the 141 fund managers who constitute the intersection of the portfolio holdingsand BEA location characteristics databases (i.e., covering the 1988-2003 period).Each manager is assigned her start-up's portfolio local bias measure. The in-dependent variables relate to the location characteristics of the CEA where themanager's firm is based.

Coefficient estimates for the regression are presented in Table 6. The neu-trality of the Home Base dummy variable indicates that start-ups whose founderschange locations are not significantly different from non-movers in their propen-sity to choose local stocks. In fact, in a separate test I find that the differencebetween the home bias measures for the two groups of firms is not distinguish-able from zero at conventional levels of statistical significance.

TABLE 6

Determinants of Net Portfolio Bias to New Location

Table 6 reports the results of fitting a regression of net portfolio bias to a start-up's new location on various ofiaraoteristicsof tfie location. To oalcuiate tfie dependent variabie. for eacfi start-up, portfoiio fioidings are extracted in tfie last quarterof tfie first year of statutory reporting of fioidings as recorded in tfie SDA/Spectrum database in tfie period 1988 to 2003.Local bias is given by tfie differenoe between tfie fraction of a start-up's totai fioidings comprising stocks located witfiin100 kiiometers of tfie firm and tfie proportion of aii stoci<s based witfiin 100 kiiometers of tfie start-up. Estabiisfiment countsand employment costs (average annual payroli per employee) are from tfie Bureau of Economic Analysis. Top InstitutionalAssets heid in eacfi CEA are from institutional Investor America's Top 300 fvioney fvlanagers surveys. Financial Oenteris an indicator of CEAs incorporating Boston. Cfiicago. Los Angeles. New York, Philadelpfiia. or San Francisco. Metro(or Non-Metro), Low Employment are 1,0 indicators of county economic typoiogies obtained from tfie U.S. Department ofAgricuiture (USDA). Tfie USDA Urban Influence code ranges from 1 (higfiest) to 12 (lowest)). Ali USDA county data areaggregated to CEA level. Home Base is a 1.0 dummy variable indicating a majority of a start-up's founders originate fromtfiat CEA. Standard errors are Wfiite heteroskedasticity-consistent. •".••. and * denote statistical significance at tfie 1%.5%, and 10% levels, respectiveiy.

Mo6el 5

Coefficient StandardEstimate Error

Code 1.0679 0.2495-Bank Establishments —0.1623 0.0503'Investment Establishments 0.0696 0.0348"Insurance Establishments 0.2516 0.0740"Top Institutional Assets 0.0099 0.0046"Bank Payroii Oosts 0.0215 0.0385Investment Payroll Costs —0.0034 0.0190Insurance Payroii Costs —0.2278 0.0642*"Metro 0.1268 0.0956Urban infiuenoe 0.0030 0.0140Financiai Center -7-0.0440 0.0264'Low Empioyment -0.1932 0.1191Population Loss 0.1793 0.1356Home Base -0.0215 0.0221Adjusted/?^ 0.10

Parwada 263

Start-ups based in areas with large numbers of investment and insurancecompanies and significant institutional assets are prone to higher levels of homebias. A one standard deviation increase in the number of investment and insuranceestablishments in a region hosting a start-up will increase its local bias by 0.07%and 0.25%, respectively. Given these rather modest magnitudes, in terms of eco-nomic significance the results are more important in giving directional guidanceon the geographic allocation investors may expect to find in start-ups' portfoliosbased on where the firms locate. The findings suggest that the exploitation of lo-cal industry networks by start-ups collocating with related industries increases thetendency to invest in familiar, local assets as proposed by Goetzmann et al. (2004).However, firms in areas populated with more banking establishments tend to re-duce their local bias, suggesting competition from bank investment managementoperations drives start-ups' security selections further away.

The hypothesis that areas with high employment costs could signal lowerlevels of local bias finds partial support in the regression results. Only insur-ance industry labor costs are negatively related to home bias. Start-ups in re-gions with more payroll costs in the investment and banking industries are proneto significant local bias. Finally, among the factors representing the social andeconomic characteristics of fund managers' locations, only the Einancial Centervariable is significant in explaining home bias. Start-ups that choose financialcenters for their locations show lower home bias. This result contradicts Covaland Moskowitz (2001) who find that home bias is higher for managers domiciledin financial centers. However, it is consistent with the expectation that start-upsdifferentiate themselves by locating their portfolios further away.

Two related themes motivated in my hypotheses regarding the determinantsof home bias remain to be examined: i) the number of fund managers, and ii) thepresence of stocks in an area. The effects of the concentration of fund man-agers in an area relative to the stocks available close to the location can be probedfurther. I define a new variable, Manager-Eirm Density as the intersection of adummy variable indicating CEAs with an above median number of stocks withina 100-kilometer radius and a dummy variable for CEAs recording below mediannumbers of investment industry establishments. This new variable is not signifi-cant and it does not significantly affect the rest of the coefficient estimates in themodel. This finding contradicts the hypothesis that entrepreneurial fund managerslocate in areas with few investment companies relative to the number of stocks.

C. The Local Bias Tendencies of Movers

The third set of tests performed in this section concerns the local bias ofmovers. As motivated in the hypothesis development section, it would be inter-esting to see how the stock selections of managers that are not constrained tomaintaining their home base on becoming entrepreneurs relate to their formerbases and their new locations, respectively. Of the 49 managers who relocateupon forming their own firms, 47 have the portfolio holdings data necessary to in-vestigate these issues. For this subset of managers, I calculate Local-Biasj^, as inequation (2) but relative, first to the new base and, second, the former base of thestart-up. Simple difference in means tests for the relative samples

264 Journal of Financial and Quantitative Analysis

are reported in Table 7. Panel A shows that the proportion of stocks held within100 kilometers of the premises of movers exceeds the value of stocks availablein the market generally within this range by 2.20%. The difference is statisticallysignificant at the 5% level. Movers do appear to hold on to stocks that were for-merly local as hypothesized. This finding suggests local bias persists for fundmanagers who relocate to form new firms.

TABLE 7

Local Bias of Portfolios of Founders That Relocate

Table 7 reports three measures of local bias in the portfolios of entrepreneurs who looate away from their origins. For eachstart-up, portfoiio hoidings are extracted in the last quarter of the first year of statutory reporting of holdings as recordedin the SOA/Spectrum database. The net locai bias, reported in Panel A as BiasJNew.Base and Bias.FormerMase, isthe difference between the start-ups' and market fractions of stocks heid iocally (within 100 kiiometers) relative to thelocation of the start-up (New Base) and the founder's former location (Former Base), respectively. Panel B comparesthe proportion of an entrepreneur's portfolio located within 100 kilometers of her former base against the proportion ofthe market portfolio iocated in that range, Panei C compares the proportion of an entrepreneur's portfoiio iocated within100 kilometers of her new base against the proportion of the market portfoiio iocated in that range, Bias.OwnJPortfoliodenotes the fraction of a start-up's total hoidings comprising stocks iocated within 100 kilometers of the location of interestand Bias-Market-Portlolio is the proportion of ali CRSP/Compustat listed stocks (or the market) that are based within 100kiiometers of the location, " and ' denote statisticai significance at the 5% and 10% ieveis, respectivelyPanel A. Net Local Bias to Former CEA Compared to New CEA

Vaiue of assets residing < 100 kilometersNumber of stocks residing < 100 kilometers

Panei B. Tendency to Invest Close to Former Base

Value of assets residing, < 100 kiiometersNumber of stocks residing < 100 kilometers

Panel C. Tendency to Invest Close to New Base

Value of assets residing < 100 kilometersNumber of stocks residing < 100 kilometers

Bias.New.Base

3,05%2,57%

Bias.Own.Portfolio

18,82%11,31%

Bias.Own.Portfolio

15,70%10,67%

BiasSormer.Base

5,25%2,63%

Bias.MarketJPortfolio

13,57%8,68%

Bias.MarketJ^ortfolio

12,65%8,10%

Difference

2,20%"0,06%

Difference

5,25%*2,63%"

Difference

3,05%2,57%

N

4747

N

4747

N

4747

In Panels B and C of Table 7,1 further investigate the sources of differencesin local bias between movers' former and current home bases. In Panel B, I re-port the tendency of movers to invest close to their new base without relating themeasure to the market portfolio as in Panel A. Likewise Panel C only consid-ers a raw measure of the tendency of movers to choose local stocks. The resultssuggest that the main driver of the difference between the net bias of movers totheir former and current bases is their tendency to retain stocks that are were for-merly local. Differences between the movers' own portfolios held locally and thestocks available in that range are statistically significant at the 10% and 5% levels,respectively, depending on whether value or number of stocks is considered. Rel-ative to the new base, the difference in managers' own portfolios and the marketportfolio available locally are not statistically significant. However, the sign onnet local bias is positive suggesting that even movers begin to tilt their portfolioselections toward their new base in the early part of their firms' existence.

In summary, the findings presented in this section show that the effects ofgeography on fund managers' portfolios are persistent. Managers who relocatewith their start-ups exhibit remnants of bias toward their prior home base whileshowing signs of a growing preference for stocks close to new locations. The

Parwada 265

beginnings of home bias at the start-up stage of the life cycle of investment com-panies studied in this paper are likely to persist as they mature.

VI. Conclusion and Suggestions for Future Research

This paper is the first attempt to examine the location decisions of start-ups inthe investment management industry. While a significant literature addresses theconsequences of location-based factors on outcomes such as the investment per-formance of portfolio managers, the origins of the location choices are not fullyunderstood. I find significant complementarities between new firm location de-cisions and the presence of investment companies, banking operations, and largeinstitutional money managers. High payroll costs are not a deterrent to location,but factors inimical to managers' personal social and economic affluence such aspoor education and employment prospects drive away start-ups. I find that a clearhome bias exists in the location decisions of entrepreneurial fund managers. Over75% of entrepreneurial fund managers locate within 25 kilometers of their formeremployers' regions. These statistics suggest that most entrepreneurial fund man-agers are constrained to their original location by such considerations as the needto perpetuate professional, social, and family networks, and relocation costs. Con-trolling specifically for start-ups whose founders mostly originate from the samearea as the new firm, I confirm that the area of origin indeed heavily conditionswhere founders locate.

The second objective of this paper is to investigate the tendencies of start-upsto bias their portfolio selections toward nearby stocks. This paper's novel contri-bution in this regard is to examine the portfolio decisions of fund managers forevidence of the influence of location attributes on stock selection tendencies. Ifind that the local bias of start-ups' equity investments is three-fold the Coval andMoskowitz (2001) measure for mutual funds. The results on the determinants oflocal bias suggest a strong correlation between the propensity to invest closer tothe home base and the presence of sub-advisory opportunities from institutionalinvestors in an area. In a final step, as a further robustness check on the role ofgeography in the genesis of home bias, I consider the portfolio holdings of man-agers who relocate when they become entrepreneurs. Consistent with the viewthat the value of local networks persists, I show that movers retain a significantbias toward stocks that were formerly local. However, nascent signs of local biastoward their new bases are also present.

The analysis presented in this paper could be extended in a number of direc-tions. First, the study offers a static view of the strategic actions of entrepreneurialfund managers at the commencement of their new firms. The portfolio strategiesfollowed by the new firms beyond the inception stage are a fruitful area for fu-ture research. Second, most founders appear to be constrained to their originalbases, implying that entrepreneurship in this industry increases local competitionfor incumbent firms. I plan to study the reaction of investment companies whosefund managers depart to form their own firms. More generally, the existence ofentrepreneurs in the investment industry presents an attractive setting in which tocarry out a detailed examination of the nature of individuals who take the self-employment career route.

266 Journal of Financial and Quantitative Analysis

References

Atmazan, A.; A. de Motta; and S. Titman. "Firm Location and ttie Creation and Utilization of HumanCapital." NBER Working Paper No. t0t06 (2003).

Bartik, T. J. "Business Location Decisions in the United States: Estimates of the Effects of Unioniza-tion, Taxes, and Other Characteristics of States." Journal of Business and Economic Statistics, 3(1985), 14-22.

Carlton, D. W. "The Location and Employment Choices of New Firms: An Econometric Model withDiscrete and Continuous Endogenous Variables." Review of Economics and Statistics, 65 (1983),440-449.

Chevatier, J., and G. Ellison. "Career Concerns of Mutual Fund Managers." Quarterly Journal ofEconomics, t t4 (t999), 389-432.

Christoffersen, S., and S. Sarkissian. "Are Fund Managers Better in Financial Centers?" WorkingPaper, McGilt University (2004).

Coval, J. D., and T. J. Moskowitz. "Home Bias at Home: Locat Equity Preference in DomesticPortfolios." Journal of Finance, 54 (t999), 2045-2073.

Coval, J. D., and T. J. Moskowitz. "The Geography of Investment: Informed Trading and AssetPrices." Journal of Political Economy, 109 (200t), 81 t-84t.

Dumais, G.; G. Ellison; and E. L. Glaeser. "Geographic Concentration as a Dynamic Process." Reviewof Economics and Statistics, 84(2002), 193-204.

Figueiredo, O.; P. Guimaraes; and D. Woodward. "Home-Hetd Advantage: Location Decisions ofPortuguese Entrepreneurs." Journal of Urban Economics, 52 (2002), 34t-361.

Fischer, J. H., and J. E. Harrington, Jr. "Product Variety and Firm Agglomeration." Rand Journal ofEconomics, 27 (1996), 281-309.

Frame, W. S., and L. J. White. "Empirical Studies of Financial Innovation: Lots of Talk, LittleAction?" Journal of Economic Literature, 42 (2004), 116-144.

French, K. R., and J. M. Poterba. "Investor Diversification and International Equity Markets." Ameri-can Economic Review Papers and Proceedings, 81 (t99t), 222-226.

Goetzmann, W. N.; M. Massa; and A. Simonov. "Portfolio Diversification and City Agglomeration."NBER Working Paper No. tO343 (2004).

Grinbtatt, M., and M. Keloharju. "How Distance, Language, and Culture Influence Stockholdings andTrades." Journal of Finance, 56 (2001), 1053-1073.

Hau, H. "Location Matters: An Examination of Trading Profits." Journal of Finance, 56 (2001),1959-1983.

Hong, H.; J. D. Kubik; and J. C. Stein. "Thy Neighbor's Portfolio: Word-of-Mouth Effects in theHoldings and Trades of Money Managers." Journal of Finance, 60 (2005), 2801-2824.

Huberman, G. "Familiarity Breeds Investment." Review of Financial Studies, 14 (2001), 659-680.Ivkovi(5, Z., and S. Weisbenner. "Local Does as Local Is: Information Content of the Geography of

Individual Investors' Common Stock Investments." Journal of Finance, 60 (2005), 267-306.Johnson, K. P. "Redefinition of the BEA Economic Areas." Survey of Current Business, 75 (1995),

75-81.Khorana, A. 'Top Management Turnover: An Empirical Investigation of Mutual Fund Managers."

Journal of Financial Economics, 40 (t996), 403^27.Khorana, A.; H. Servaes; and P. Tufano. "Explaining the Size of the Mutuat Fund Industry around the

World." Journal of Financial Economics, 78 (2005), 145-185.Lewis, K. K. "Puzzles in International Financial Markets." In Handbook of International Economics,

Vol. 3, G. Grossman and K. Rogoff, eds. Amsterdam, The Netherlands: North Holland (1995).Lewis, K. K. "Trying to Explain Home Bias in Equities and Consumption." Journal of Economic

Literature, 37 (1999), 571-608.Liang, K.-Y., and S. L. Zeger. "Longitudinal Data Analysis Using Generalized Linear Models."

Biometrika, 73 (1986), 13-22.McFadden, D. L. "Conditional Logit Analysis of Qualitative Choice Behavior." In Frontiers in Econo-

metrics, P. Zarembka, ed. New York, NY: Academic Press (1974).Porter, M. E. The Competitive Advantage of Nations. New York, NY: Free Press (1990).Shapira, Z., and I. Venezia. "Patterns of Behavior of Professionally Managed and Independent In-

vestors." Journal of Banking and Finance, 25 (2001), 1573-1587.Syverson, C. "Market Structure and Productivity: A Concrete Example." Journal of Political Econ-

omy, 112(2004), 1181-1222.Villalonga, B. "Diversification Discount or Premium? New Evidence from the Business Information

Tracking Series." Journal of Finance, 59 (2004), 479-506.Zingales, L. "In Search of New Foundations." Journal of Finance, 55 (2000), 1623-1653.