No Place Like Home: Familiarity in Mutual Fund Manager Portfolio Choice Veronika K. Pool Indiana University Noah Stoffman Indiana University Scott E. Yonker Indiana University We show that familiarity affects the portfolio decisions of mutual fund managers. Controlling for fund location, funds overweight stocks from their managers’ home states by 12% compared with their peers. In team-managed funds, home-state overweighting is 37% larger than the fund location effect. The home-state bias is stronger if the manager is inexperienced, is resource-constrained, or spent more time in his home state. Home-state stocks do not outperform other holdings, confirming that home-state investments are not informed. The overweighting also leads to excessively risky portfolios. (JEL G11, G23)

Investors invest in what they know: a “home bias” in equity holdings is pervasive around the world (French and Poterba 1991; Tesar and Werner 1995; Kang and Stulz 1997), and domestic portfolios are also often tilted toward nearby firms (Coval and Moskowitz 1999; Grinblatt and Keloharju 2001; Ivkovic´ and Weisbenner 2005; Seasholes and Zhu 2010). Whether the local equity preference of individual investors is driven by information or familiarity is a subject of debate in the literature.1 Among professional investors, however, it would be surprising if a preference for local stocks were not driven by information, since these investors face higher costs from poor portfolio performance due, for example, to career concerns (Chevalier and

For helpful comments, we thank Robert Battalio, Nicole Boyson, Utpal Bhattacharya, David Hirshleifer (the editor), Byoung-Hyoun Hwang, Andrew Karolyi, Berk Sensoy, Tyler Shumway, Michael Weisbach, Scott Weisbenner, and Rohan Williamson, as well as seminar participants at the Stockholm School of Economics, Maastricht University, and participants of the Conference on Financial Economics and Accounting, the IUNotre Dame-Purdue Summer Symposium, the Indiana University finance brown bag, and the Ohio State University Finance Alumni Conference. The authors are grateful for funding from the Kelley School of Business Matching Grant program at Indiana University. Send correspondence to Scott Yonker, Kelley School of Business, Indiana University, 1309 E 10th Street, Bloomington, IN, 47405; telephone: (812) 855-2694. E-mail: [email protected]. 1 Ivkovi´c and Weisbenner (2005) find support for the information hypothesis, while Grinblatt and Keloharju (2001),

Huberman (2001), and Seasholes and Zhu (2010) find evidence that investors suffer from a familiarity bias. © The Author 2012. Published by Oxford University Press on behalf of The Society for Financial Studies. All rights reserved. For permissions, please e-mail: [email protected]. doi:10.1093/rfs/hhs075

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2563

2563–2601

The Review of Financial Studies / v 25 n 8 2012

Ellison 1999b). Consistent with this, Coval and Moskowitz (2001) find support for informed local investing by mutual funds. In this article, we ask whether mutual fund managers also exhibit a familiarity bias in their portfolio allocation decisions. The challenge in testing whether familiarity plays a role in the portfolio choice of professional managers is to identify securities that are, ex ante, likely to be familiar to the fund manager, but for which the manager has no informational advantage. We hypothesize that firms headquartered in the fund manager’s home state satisfy these criteria.2 For example, we argue that a manager who grew up in Indiana and who works for a fund in New York is likely to be familiar with Indiana-based firms, but not necessarily informed about them. Our motivation for this identification strategy is twofold. First, in contrast to the current location of the mutual fund, it is less obvious that the manager has active ties to his home state. Second, the fund’s only connection to the home state is through the manager; most of the fund’s “information gatherers” do not have a connection to the state. Whether securities from managers’ home states are merely familiar is an empirical question. Therefore, to investigate the role of familiarity in portfolio decisions, we proceed in two steps. First, we ask whether managers overweight companies from their home states. Second, we ask whether these choices reflect information by examining the returns of home-state investments. Under the information hypothesis, managers may have access to information about companies at home. For example, friends and family may still live in the area and work for nearby companies. These contacts can provide information about business conditions at their employers, even if they rarely provide any illegal inside information. The value of the information need not be especially high—as long as the manager has a comparative advantage in gathering information from his home state, he or she will be more likely to trade on it. If managers are informed about home companies, we would expect them either to overweight or underweight these companies—depending on whether their information is positive or negative—and generate positive abnormal returns from their information. In contrast, under the familiarity hypothesis, managers may simply be more familiar with home-state companies, even if they have no real information about them. While professional fund managers are probably aware of most, if not all, stocks in their investment universe, if managers are more familiar with firms from their home states, then perhaps when choosing among stocks they choose the familiar one. There are several reasons why managers might choose stocks with which they are merely familiar. First, they may feel a connection to their home states because of “homophily” (Lazarsfeld and Merton 1954), so investing in home-state companies may simply feel good. Second, managers 2 Throughout the article, we refer to the state where the manager grew up as their “home state.” We identify home

states using Social Security data, as explained later.

2564

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2564

2563–2601

No Place Like Home

may be more pessimistic about the prospects of unfamiliar companies.3 Third, managers may falsely believe that their connections to their home state provide them with a comparative advantage, even if they do not. This type of perceived information advantage would likely involve less well-known stocks, as we discuss later. The empirical predictions of the familiarity hypothesis differ starkly from those of information. If managers are simply more familiar with companies from home, they should overweight only home-state companies. Furthermore, the performance implications of familiarity depend on whether managers are skilled in general. If managers have no talent in picking stocks, relying on familiarity to construct the portfolio will have no performance effects since familiarity is no better or worse than any of their other uninformed methods of selecting stocks. In contrast, when otherwise skilled managers suffer from a familiarity bias, overinvesting in home-state stocks should lower their returns since the home-state portion of the portfolio is built on biased—rather than informed—choices. We find that mutual fund managers invest more in stocks headquartered in the state where they grew up than do peers of similar funds who grew up elsewhere: funds invest 12% more in their managers’ home states than would otherwise be expected. We also show that around manager turnovers, new managers build up their holdings in home-state stocks within a few quarters after their arrival. Our findings are robust to alternative geographic classifications and distance measures, and to including fund fixed effects, which identifies the effect using within-fund variation and thus presents a very high econometric hurdle. In relation to the extant literature, we provide a direct comparison between the role of manager home states and that of the fund’s current location in portfolio allocation decisions (Coval and Moskowitz 1999). Surprisingly, where a fund’s managers grew up is 70% as important as where the fund is currently located in determining its portfolio picks. Moreover, when a fund is team-managed, the abnormal weight on home-state securities is 37% higher than that of companies located close to the fund.4 The magnitude of our results is also similar to the effects of political values (Hong and Kostovetsky 2012) and of managerdirector college networks (Cohen, Frazzini, and Malloy 2008) on mutual fund portfolio choice. Our cross-sectional analyses lend additional support to the familiarity hypothesis. Managers put more weight on home-state companies when they are early in their careers, suggesting that they have not yet fully developed ways of gathering information. Additionally, home-state overweighting is stronger

3 Strong and Xu (2003) find in survey data that fund managers hold pessimistic beliefs toward unfamiliar stocks.

Cao, Han, Hirshleifer, and Zhang (2011) model investment decisions when investors generally prefer the status quo due to this type of pessimism. 4 This is consistent with what industry professionals have told us about common practice: each manager contributes

stock picks to the portfolio, and as a result, the cumulative effect of the home-state tilt is large.

2565

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2565

2563–2601

The Review of Financial Studies / v 25 n 8 2012

among managers who spent more time in their home states and those of funds with fewer resources. Consistent with a perceived information advantage in less well-known securities, we also find that home-state overweighting is more pronounced in small stocks, and in those that are not included in the S&P 500, have fewer sales, are followed by fewer analysts, and have low advertising expenditures. We then investigate fund performance. We divide each fund’s portfolio into home-state holdings, local holdings (i.e., stocks headquartered in the state where the fund is currently located), and the remainder. The rationale for partitioning non-home-state holdings is motivated by the literature on mutual fund manager skill. While many studies find that managers have no skill in general,5 recent articles argue that some holdings may still be informed.6 To isolate those portfolio stocks that are most likely to be informed, we rely on Coval and Moskowitz (2001), who argue that local stocks compose such a portfolio segment. We find strong empirical support for the familiarity hypothesis: A fund’s holdings in its managers’ home-state firms do not outperform its other holdings. More specifically, home-state holdings underperform nearby investments (i.e., underperform the fund’s “local” portfolio), but perform similarly to the rest of the fund’s portfolio. Finally, we document that one cost of home-state overweighting is excess risk. In particular, we show that idiosyncratic volatility is greatest among managers who overweight the most. The underdiversification is not driven mechanically by biased funds holding fewer stocks; rather, funds that exhibit a larger home-state tilt have higher idiosyncratic volatilities than other funds with the same number of stock holdings. This result is consistent with Pirinsky and Wang (2006), who show that the returns of stocks headquartered in the same area exhibit substantial comovement. Our finding suggests that mutual funds whose managers exhibit the largest familiarity biases hold inefficient portfolios. Our article contributes to the active literature showing that the portfolio decisions of both individual and institutional investors are affected by their experiences and values. A wide variety of investor characteristics, including age, past trading experience, place of employment, and political attitudes, have all been found to affect portfolio decisions.7 Attitudes about countries have similar effects.8 Investors have also been shown to tilt their portfolios toward

5 See, for example, Jensen (1968), Gruber (1996), Chevalier and Ellison (1999a), and Carhart (1997). 6 These include Coval and Moskowitz (2001), Alexander, Cici, and Gibson (2007), Cremers and Petajisto (2009),

Pomorski (2009), and Cohen, Polk, and Silli (2010). 7 See Korniotis and Kumar (forthcoming), Greenwood and Nagel (2009), Goetzmann and Kumar (2008), Seru,

Shumway, and Stoffman (2010), Malmendier and Nagel (2011), Cohen (2009), and Hong and Kostovetsky (2012). 8 See Bhattacharya and Groznik (2008), Hwang (2011), and Morse and Shive (2011).

2566

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2566

2563–2601

No Place Like Home

companies that are geographically close to them and companies that are held by their neighbors.9 Whether this tilt reflects information is the subject of some debate. Among individual investors, Ivkovic´ and Weisbenner (2005) find higher returns on local holdings than non-local holdings, although Seasholes and Zhu (2010) draw contradictory conclusions using different performance evaluation measures. Coval and Moskowitz (1999, 2001) show that fund managers invest more in companies whose headquarters are located close to the fund, and earn higher returns on their local holdings. Consistent with this result, Parwada (2008) shows that startup funds frequently locate close to the founder’s previous employer and that these funds invest heavily in local stocks. Similarly, Cohen, Frazzini, and Malloy (2008) identify networks of fund managers and board members based on college attendance and find that fund managers earn higher returns on firms to which they are connected. We contribute to this literature by using a unique identification strategy to show that fund managers exhibit familiarity in their portfolio choice. We overcome a challenge faced by previous studies by disentangling the effect of the manager from that of the fund. Our results are consistent with a behavioral explanation for the preference for home-state securities, whereby less-sophisticated managers use familiar stocks as their investment ideas. 1. Data and Sample Construction We combine several data sources in this study. First, we draw information on fund managers from Morningstar, which reports the name of each manager for a fund (including individuals on team-managed funds), their start and end dates with the fund, and information about the manager’s educational background. We limit our sample to actively managed U.S. equity funds by filtering the observations using Morningstar style categories as well as manually screening the fund names. Our second data source is the Thomson Financial CDA/Spectrum Mutual Fund database. The database contains the quarter-end holdings reported by U.S.-based mutual funds in mandatory Securities and Exchange Commission filings. Thomson uses two date variables, RDATE and FDATE, which refer to the actual date for which the holdings are valid and the Thomson vintage date on which the data were cut, respectively. We follow standard practice and restrict the holdings to those observations where the FDATE is equal to the RDATE to avoid the use of stale data in our analysis. From this starting point, there are 4,731,878 quarterly fund-holding observations from the first quarter of 1996 to the fourth quarter of 2009.

9 See Coval and Moskowitz (1999, 2001), Grinblatt and Keloharju (2001), Huberman (2001), Hong, Kubik, and

Stein (2004), and Shive (2010).

2567

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2567

2563–2601

The Review of Financial Studies / v 25 n 8 2012

Since we are interested in the domestic portion of the funds’ portfolios, we remove holdings in firms headquartered outside the United States. This reduces the number of quarterly fund holding observations to 4,446,768. There are 48,077 quarterly fund observations with 2,552 distinct funds holding 13,908 different securities.10 In addition, we restrict the sample to those funds with a Morningstar category in the 3-by-3 size/value grid (US Large Blend, US Large Growth, US Large Value, US Mid-Cap Blend, US Mid-Cap Growth, US Mid-Cap Value, US Small Blend, US Small Growth, or US Small Value). The rationale for this restriction is twofold: First, by restricting the sample to these categories, we are able to create benchmark portfolios that are both broad enough for the weights of these portfolios to be precisely measured and narrow enough to be relevant. Second, using only these broad categories mitigates the concern that any geographic overweighting in our sample may be driven by industry-specific funds, since these funds concentrate their holdings in certain industries, and firms within industries tend to cluster geographically. This filter further reduces the sample to 42,109 quarterly fund observations with 4,236 managers and 2,192 distinct funds. When constructing the weights in our benchmark portfolios, this sample of 42,109 quarterly fund observations is used. We collect data on the managers’ home states from the Lexis Nexis Online Public Records Database following the methodology proposed by Yonker (2010), who uses the first five digits of Social Security Numbers to identify the state and year in which CEOs grew up. (We provide details of the procedure in the appendix.) We are able to determine the home state for 2,143 of the 4,236 unique managers in our restricted sample.11 This reduces the sample to 27,914 quarterly fund observations with 1,810 unique funds.12 Finally, we remove mutual funds for which the location of the fund complex headquarters is missing. This final filter produces a sample of 27,430 quarterly fund observations (43,632 quarterly manager-fund observations) with 1,767 unique funds managed by 2,109 unique managers. In addition to the data sources used in the sample construction, we locate the management company’s address and obtain fund returns from the CRSP Survivor-Bias-Free Mutual Fund database. We use fund N-SAR filings to find the address of each fund’s advisers and subadvisers during our sample period (see the appendix for more details). Compustat and Compact Disclosure are the sources of headquarters locations for each stock held in the mutual fund 10 Approximately 97% of the value of all holdings remains after this filter. 11 The percentage of mutual fund managers identified is considerably lower than that in Yonker (2010), because,

unlike Execucomp, Morningstar does not provide precise data on an individual’s age. Adding an age restriction greatly increases the probability of identifying a unique individual. 12 Table W1 of the Online Appendix shows that the sample of funds for which managers are identified and those

for which they are not are very similar. The exception is that funds for which managers are identified tend to be larger, but this biases our results against finding home-state overweighting, since we show later that larger funds have smaller home-state biases.

2568

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2568

2563–2601

No Place Like Home

Table 1 Sample composition and summary statistics Panel A: Sample composition Morningstar Category

Sample avg. aggregate TNA per quarter ($ millions)

Sample avg. fund TNA per quarter ($ millions)

Sample avg. Funds per quarter

Avg. pct. of benchmark TNA covered per quarter

Avg. pct. of benchmark funds covered per quarter

US Large Blend US Large Growth US Large Value US Mid-Cap Blend US Mid-Cap Growth US Mid-Cap Value US Small Blend US Small Growth US Small Value

126,757 189,972 135,820 13,242 36,266 15,609 19,679 20,892 8,312

1,496 1,654 1,864 471 652 684 595 340 375

86 116 72 29 57 18 31 61 20

82.91 84.04 84.02 72.19 65.62 67.15 82.51 72.79 80.41

68.32 71.85 72.09 70.02 63.57 67.34 68.31 69.08 75.07

Total

566,549

1,157

490

81.03

69.49

Panel B: Summary statistics Variable

Mean

Median

SD

N

Fund total net assets ($ billions) Fund family total net assets ($ billions) Fund age (FundAge) Growth fund dummy (Growth) Value fund dummy (Value) Small-cap fund dummy (SmallCap) Large-cap fund dummy (LargeCap) Subadvised dummy (Subadvised) Number of managers Team managed (TeamManaged) Manager age (Age) Manager tenure Manager experience (Experience) Manager home tenure (HmTenure) Attended an Ivy League school (IvyLeague) Manager from same state as fund Manager from same state as college Distance from manager home state to fund location (miles)

1.16 65.77 8.83 0.48 0.22 0.23 0.56 0.32 2.45 0.65 47.01 4.75 8.25 34.70 0.19 0.29 0.44 851.83

0.18 8.69 6.01 0.00 0.00 0.00 1.00 0.00 2.00 1.00 45.00 3.21 6.75 31.00 0.00 0.00 0.00 602.68

4.70 186.44 9.89 0.50 0.42 0.42 0.50 0.47 2.12 0.48 10.94 4.93 6.44 14.58 0.39 0.45 0.50 784.83

27,430 24,205 24,960 27,430 27,430 27,430 27,430 27,430 27,430 27,430 40,752 43,632 43,632 20,671 36,723 43,632 28,097 30,923

Panel A of the table reports the average aggregate total net assets (TNA), the average fund’s TNA, the average number of funds in the sample, the average percentage of aggregate TNA of the benchmark covered, and the average percentage of benchmark funds covered per quarter by Morningstar category for the sample of 27,430 quarterly fund observations from the first quarter of 1996 through the fourth quarter of 2009. Panel B reports summary statistics for fund and manager characteristics for the sample. For fund-specific variables, the unit of observation is fund-quarter, and for manager-specific variables, the unit of observation is fund-manager-quarter. The sample includes 27,430 quarterly fund observations and 43,632 quarterly manager observations.

portfolios. Stock characteristics are from Compustat, and analyst coverage is from I/B/E/S. When a college name is available in Morningstar, we determine its location with an Internet search. Panel A of Table 1 shows the average quarterly composition of the sample by Morningstar fund category. There are 490 funds included in the sample on average each quarter. This represents 69.5% of the the funds and 81.0% of total net assets (TNA) in the Morningstar benchmark funds. The largest Morningstar category represented, by both number of funds and TNA, is Large Growth, with on average 116 funds in the sample each quarter and with average aggregate

2569

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2569

2563–2601

The Review of Financial Studies / v 25 n 8 2012

TNA of $190 billion. The category with the fewest funds is Mid-Cap Value with an average of 18 funds, while Small Value is the smallest category, with an average quarterly aggregate TNA of $8.3 billion. In Panel B of the table, we report summary statistics for fund- and managerspecific variables. The average (median) fund in the sample has TNA of $1.16 billion ($0.18 billion); 47% of the funds in the sample are growth funds, 56% are large-cap funds, and 32% are subadvised. The median fund is managed by two managers, and 65% of funds employ more than one manager in a given quarter. The median manager is 45 years old, has been a portfolio manager for 6.8 years, and lived in his home state for the first 31 years of his life. Approximately 19% of manager-quarters have managers with a degree from an Ivy League institution. To display the geographic dispersion of managers and funds, we count the number of active funds located in a state each quarter during our sample, as well as the number of fund managers who come from each state. We report the time-series average of these counts in Figure 1. Not surprisingly, both the location of funds and the home states of fund managers are correlated with population distribution. Many funds are concentrated in New York, California, and Illinois, but funds are also strongly represented in Colorado, Connecticut, Florida, Massachusetts, Minnesota, Texas, and Wisconsin, among others. Fund managers are also likely to be born in these states, but home states are fairly dispersed and at least one manager comes from each state.

Figure 1 Fund locations and manager home states The figure shows the average number of funds located in each state (indicated by state shading), and the average number of fund managers who grew up in each state (indicated by bars). Data are calculated each quarter for each state, and then averaged across time. Alaska and Hawaii are displayed below the lower 48 states. An average of less than one means that a state is not represented among active managers or active funds during some part of our sample period.

2570

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2570

2563–2601

No Place Like Home

2. Do Funds Overweight Stocks from Their Managers’ Home States? If mutual fund managers exhibit a familiarity bias toward companies in their home states, then we should expect that they overweight these companies in their portfolios. Superior information could also lead managers to home-state overweighting, but it could also lead to underweighting if the information is negative. We begin our empirical analysis by examining the portfolio weight that fund managers place on their home states. We do so by estimating various forms of the regression equation wi,s,t = βPctMgrHmStatei,s,t +δMorningstarBMWti,s,t +  Controlsi,s,t +i,s,t ,

(1)

where wi,s,t is the weight in fund i’s portfolio of firms headquartered in state s during quarter t, PctMgrHmStatei,s,t is the ratio of the number of managers of fund i from state s to the total number of managers of fund i during quarter t, MorningstarBMWti,s,t is the average portfolio weight in state s of all funds within the same Morningstar category as fund i during quarter t, and Controlsi,s,t is a vector of relevant control variables. If fund managers tilt their portfolios toward their home states, then we should find that β is positive and statistically significant. Our estimate of β is the total home-state bias across all managers of the fund, thus representing the percentage of the average fund’s assets abnormally allocated to its managers’ home-state firms. If all funds in the sample had only one manager, then β would also measure the average home-state bias per manager. If instead each fund has two managers, and on average each manager allocates an extra 50 basis points (bps) to companies in his home state, then β would be 1.00. An alternative to estimating (1) using quarterly fund-state observations is to use quarterly fund-stock observations. Although we estimate such a regression later when investigating in which types of stocks home-state overweighting is the largest, in our base specification we aggregate stocks to the state level. We do this for several reasons. First, the article is about overweighting in home states. Therefore, it seems natural to estimate the abnormal portfolio weight at the state level. Additionally, estimating a fund-stock regression can only tell us how much each home-state stock is overweighted on average, but not the total portion of the portfolio abnormally allocated to home-state stocks, which is the economically relevant number. In Table 2, we report the coefficient estimates and standard errors clustered at the fund level from the OLS estimation of various forms of regression (1). In column 1, we include only PctMgrHmState and a constant in the regression. The average weight that managers at a fund invest in their home states is given by the sum of PctMgrHmState and the intercept, so we estimate that 8.46% of

2571

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2571

2563–2601

[10:29 2/7/2012 OEP-hhs075online.tex]

0.02 1,392,606 No

1.87∗∗∗ (0.00)

1.34∗∗∗ (0.20)

6.59∗∗∗ (0.31)

0.70 1,392,606 No

0.99∗∗∗ (0.01) 0.00 (0.01)

Full (2)

Full (1)

0.70 1,392,606 No

0.98∗∗∗ (0.01) −0.01 (0.01)

0.87∗∗∗ (0.18) 1.23∗∗∗ (0.16)

Full (3)

0.68 707,370 No

0.97∗∗∗ (0.01) 0.01 (0.01)

0.82∗∗∗ (0.19) 1.53∗∗∗ (0.23)

Hm found for all mgrs (4)

0.68 483,837 No

0.98∗∗∗ (0.01) 0.00 (0.02)

0.58∗∗∗ (0.19) 1.57∗∗∗ (0.25)

Single manager (5)

0.70 908,769 No

0.99∗∗∗ (0.01) −0.01 (0.01)

1.39∗∗∗ (0.28) 1.01∗∗∗ (0.18)

Team managed (6)

0.70 1,392,606 No

0.70∗∗∗ (0.14) 1.12∗∗∗ (0.16) 0.65 (0.53) 0.98∗∗∗ (0.01) 0.00 (0.01)

Full (7)

0.85 1,392,606 fund-state

0.91∗∗∗ (0.02) 0.17∗∗∗ (0.04)

0.48∗∗ (0.20)

Full (8)

where wi,s,t is the weight in fund i ’s portfolio of firms headquartered in state s during quarter t , PctMgrHmStatei,s,t is the ratio of the number of managers of fund i from state s to the total number of managers of fund i during quarter t , MorningstarBMWti,s,t is the average portfolio weight in state s of all funds within the same Morningstar category as fund i during quarter t , and Controlsi,s,t is a vector of relevant control variables. If fund managers tilt their portfolios toward their home states, then we should find that β is positive and significant. The sample includes 1,392,606 quarterly fund-state observations from the first quarter of 1996 to the fourth quarter of 2009 and includes 1,767 unique funds managed by 2,109 unique managers. MFHQStatei,s,t is a dummy variable that takes a value of one if the mutual fund complex of fund i is headquartered in state s during quarter t and is zero otherwise. In column 4, the sample is limited to funds whose managers’ home state was identified for all fund managers. Columns 5 and 6 include funds with only a single manager and only those that are team-managed, respectively. In column 8, we include fund-state fixed effects. Standard errors, clustered at the fund level, are in parentheses. Significance levels are denoted by *, **, and ***, which correspond to the 10%, 5%, and 1% levels, respectively.

wi,s,t = β PctMgrHmStatei,s,t +δ MorningstarBMWti,s,t +  Controlsi,s,t +i,s,t ,

The table reports the coefficient estimates and standard errors from the OLS estimation of various forms of the regression equation

AdjR2 N Fixed effects

Intercept

MorningstarBMWt

PctMgrHmState × MFHQState

MFHQState

PctMgrHmState

Sample:

Table 2 Do funds overweight stocks from their managers’ home states?

The Review of Financial Studies / v 25 n 8 2012

2572

Page: 2572

2563–2601

No Place Like Home

mutual fund portfolios are allocated to firms headquartered in the home states of their managers. In column 2, we add MorningstarBMWt to control for the average portfolio weight that funds in the same Morningstar category allocate to a given state during each quarter. The estimate on MFHQState is close to one and highly significant. The inclusion of this benchmark also shrinks the intercept to zero and explains much of the variation in portfolio weights across funds, suggesting that we are using the right benchmark. After controlling for the benchmark weight, the coefficient estimate on PctMgrHmState is 1.34 and is significant at greater than the 1% level. Since β measures a per fund effect, this indicates that the average fund in the sample overweights stocks from its managers’ home states by 134 bps compared with other funds in the same nine-box Morningstar category. Columns 1 and 2 therefore imply that the expected state weight in the absence of a home bias is 8.46−1.34 = 7.12%, and that the average fund overweights its managers’ home states by 134/712 = 18.8%. If labor markets for fund managers are geographically segmented, then funds might be more likely to hire managers who are close to the fund headquarters. This suggests that a portion of home-state overweighting could be due to the findings of Coval and Moskowitz (1999), who show that mutual funds overinvest in stocks that are headquartered nearby. To check this, in column 3 we control for the fund’s location by adding a dummy variable that takes a value of one if the mutual fund complex of fund i is headquartered in state s during quarter t (MFHQState). The coefficient estimate on MFHQState implies that funds overweight stocks from their state by 123 bps relative to funds in the same nine-box Morningstar category. The coefficient on PctMgrHmState decreases, but remains highly statistically significant. The regressions thus far include all fund observations where the home state of at least one manager was identified. In column 4, we estimate the regression from column 3 for the sample of observations for which we identify the home state of all managers. The estimate on PctMgrHomeState in column 4 is 82 bps, which is very similar in magnitude to the estimate in column 3. Therefore, our results do not seem to be biased by not identifying each manager’s home state. While the average fund overweights the home states of its managers, our sample contains both single-managed and team-managed funds. It is interesting to see whether the abnormal portfolio tilt of each manager simply adds up in a team-managed fund or the presence of multiple managers mitigates the biases of the individuals. The answer to this question provides some insights into how manager teams operate. In columns 5 and 6, we estimate the regression from column 3 for the two subsamples. In both groups, we find significantly positive home-state biases (at the 1% level). Single-managed funds overweight their home states by 58 bps, while team-managed funds overweight their managers’home states by 139 bps. Therefore, although the presence of multiple decision-makers in team-managed funds appears to mitigate the overweighting

2573

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2573

2563–2601

The Review of Financial Studies / v 25 n 8 2012

to some extent, the home-state bias, in absolute magnitude, aggregates in the portfolio. In column 7 of the table, we orthognalize PctMgrHmState to MFHQState by including the interaction of the two in the regression. The coefficient on PctMgrHmState remains statistically significant at 70 bps. Finally, in column 8 of the table, we estimate our model with fund-state fixed effects. This specification controls for the average weight each fund has in each state, so the estimate of β is identified from within-fund variation in manager home states. The coefficient estimate on PctMgrHmState is 48 bps and remains statistically significant. How does the home-state bias compare with other portfolio effects documented in the literature? First, our regression framework allows us to evaluate the relative importance of manager home states in the average fund’s overall portfolio, compared with that of the current location of the fund (Coval and Moskowitz 1999). Column 3 reports that the cumulative home-state tilt is approximately 87 bps for the average fund in our sample, and is roughly 70% as large as local overweighting, which is 123 bps. Moreover, for team-managed funds, the absolute magnitude of home-state overweighting is 37% larger than that of local overweighting. In particular, for these funds, home-state stocks receive an additional 1.39% weight in the overall portfolio allocation, while the abnormal weight on local stocks is 1.01%. Second, the magnitude of the home-state tilt is also similar to the effects of political values and of fund manager–corporate director college networks on mutual fund portfolio choice. Hong and Kostovetsky (2012) show that Democratic managers underweight politically sensitive industries by 68 bps, which is about 19% of the average weight of these industries in mutual fund portfolios. Cohen, Frazzini, and Malloy (2008) estimate that mutual fund managers overweight stocks in which they have college connections with board members by between 10% and 14%, but do not report the percentage of the average fund’s portfolio that is allocated to such investments.13 Since these figures are not adjusted for the fund location effect, our estimate in column 2 provides the most direct comparison. The absolute magnitude of the abnormal home-state weight in this column is 134 bps. This corresponds to about 18.8% home-state overweighting relative to the portfolio choices of unbiased funds. 2.1 Changes in overweighting around manager turnover In Table 2, we show that funds overinvest in their managers’ home stocks using a regression framework. To establish an even cleaner link, we next investigate changes in portfolio allocations around manager turnover. If managers bias fund holdings toward home states, then we should find that new managers increase

13 These estimates come from the specifications that include control variables in Table 2 of Cohen, Frazzini, and

Malloy (2008).

2574

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2574

2563–2601

No Place Like Home

Table 3 Home-state overweighting around manager turnover

Excess weight in former manager’s home state Excess weight in new manager’s home state

Prior to turnover

Following turnover

Difference

0.48∗∗ (0.23) 0.29 (0.19)

0.29 (0.23) 0.65∗∗∗ (0.20)

−0.19 (0.17) 0.36∗∗ (0.17)

The table reports mutual funds’ average excess weights in their former and new managers’ home states one year prior and one year following manager turnover, as well as the difference in excess weights. Excess portfolio weights are calculated as a fund’s actual portfolio weight in its manager’s home state minus the portfolio weight of its Morningstar benchmark in that state. The analysis uses 322 fund manager turnover events from 1996 to 2009 where the former and new managers come from different home states. In cases where one manager is replaced by several managers or one manager replaces several managers, the excess weights of each of the managers’ home states are averaged. Stars on the excess weights indicate the significance level of the t-test testing whether the estimate is significantly different from zero. Standard errors are reported in parentheses. Significance levels are denoted by *, **, and ***, which correspond to the 10%, 5%, and 1% levels, respectively.

the fund’s allocation to their home states, while they may also decrease holdings in the previous managers’ home states. Table 3 displays mutual funds’ average excess portfolio weights on companies in their former and new managers’ home states one year prior and one year following manager turnover. The excess portfolio weights are calculated as a fund’s actual portfolio weight in its manager’s home state minus the porfolio weight of its Morningstar benchmark in that state. The table shows that while funds significantly overweight their current manager’s home state prior to turnover, they do not significantly overweight the incoming manager’s home state. However, once the new manager is actually managing the fund, the excess portfolio weight in his home state increases by 36 bps (significant at the 5% level).14 Table 3 also shows that the increase in the abnormal weight allocated to the new manager’s home state is greater than the decrease in that of the previous manager. This result is similar to Cohen, Frazzini, and Malloy (2008), who also document an asymmetry in portfolio weight changes around manager turnovers. Finally, the total effect of the turnover is the difference-in-differences estimate. This is given by the difference between the changes in the excess weights reported in the last column of the table, and corresponds to 55 bps (significant at the 1% level). The magnitude of this estimate is consistent with that reported for single-manager funds in Table 2 (58 bps), which is not surprising since the difference methodology measures the home-state effect on a per manager basis. 2.2 Fund characteristics To gain further insights into what drives funds to overweight their managers’ home states, we next investigate which types of funds overweight the most. 14 These findings are similar in spirit to those of Jin and Scherbina (2011), who find that new managers sell their

predecessors’ losing stocks faster than stocks in other momentum portfolios, or than continuing managers.

2575

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2575

2563–2601

The Review of Financial Studies / v 25 n 8 2012

Table 4 Fund characteristics and home-state overweighting

PctMgrHmState PctMgrHmState × Value PctMgrHmState × Growth

(1)

(2)

(3)

(4)

(5)

(6)

1.00∗∗ (0.46) −0.20 (0.56) −0.18 (0.51)

1.55∗∗∗ (0.44)

2.28∗∗∗ (0.78)

1.13∗∗∗ (0.28)

0.67∗∗∗ (0.19)

2.75∗∗ (1.14) −0.33 (0.60) −0.29 (0.56) −0.49 (0.51) −0.64 (0.55) −0.51∗∗ (0.26) 0.15 (0.16) 0.62∗∗ (0.29) 1.20∗∗∗ (0.17) 0.98∗∗∗ (0.01) 0.69 1,233,945

PctMgrHmState × SmallCap

−0.79 (0.50) −0.88∗ (0.52)

PctMgrHmState × LargeCap PctMgrHmState × FamSizeQuin

−0.43∗∗ (0.20)

PctMgrHmState × FundSizeQuin

−0.13 (0.11) 1.23∗∗∗ (0.16) 0.98∗∗∗ (0.01)

1.23∗∗∗ (0.16) 0.98∗∗∗ (0.01)

1.24∗∗∗ (0.17) 0.98∗∗∗ (0.01)

1.23∗∗∗ (0.16) 0.98∗∗∗ (0.01)

0.62∗∗ (0.27) 1.21∗∗∗ (0.16) 0.98∗∗∗ (0.01)

0.70 1,392,606

0.70 1,392,606

0.69 1,233,945

0.70 1,392,606

0.70 1,392,606

PctMgrHmState × TeamManaged MFHQState MorningstarBMWt AdjR2 N

The table reports the coefficient estimates and standard errors from the OLS regression equation estimated in column 3 of Table 2, including interaction terms with various fund characteristics. Observations are quarterly fund-state observations. Value is a dummy variable that is one if the mutual fund is categorized by Morningstar as a value fund. Growth is a dummy variable that is one if the mutual fund is categorized by Morningtar as a growth fund. SmallCap is a dummy variable that is one if the mutual fund is categorized by Morningtar as a smallcap fund. LargeCap is a dummy variable that is one if the mutual fund is categorized by Morningtar as a large-cap fund. FundSizeQuin is equal to the fund’s TNA quintile minus one, where one is the smallest quintile based on total net assets of the fund each quarter. FamSizeQuin is equal to the fund’s fund family TNA quintile minus one, where one is the smallest quintile based on total fund family net assets each quarter. TeamManaged is a dummy variable that is one if the fund is managed by more than one manager. All specifications include a constant and the main effect for the interaction variables being tested, although for brevity we do not report these coefficient estimates. Standard errors, clustered at the fund level, are in parentheses. Significance levels are denoted by *, **, and ***, which correspond to the 10%, 5%, and 1% levels, respectively.

Specifically, we test whether there are differences in overweighting across funds by investment style, fund and fund-family resources, and the structure of the management team. We start by testing for differences in overweighting across fund investment styles by interacting PctMgrHmState with dummy variables that indicate the Morningstar styles of the funds: value, growth, small-cap, and large-cap. If there are differences in manager home-state weightings across styles, then we should find that these interaction terms are significantly different from zero. Table 4 shows the regression results for our tests. The baseline model used in the table is from column 3 of Table 2. In column 1, we test for differences in manager home-state weightings across value, growth, and blend funds. The coefficient estimates on the PctMgrHmState interactions with Value and Growth are not statistically different from zero, indicating that there is no difference in the weight that managers place on their home states across these funds. In column 2, we test for differences across size investment objectives of funds. Large-cap funds have the lowest home-state biases, followed by small-cap and

2576

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2576

2563–2601

No Place Like Home

then mid-cap funds, but the only significant difference between these fund styles is between large-cap and mid-cap funds. Smaller funds and those from smaller fund families are likely to have fewer resources to devote to investment analysis, and thus are more prone to rely on their managers’ideas (and consequently, their biases). We find that this is indeed the case. In column 3, we test for differences in the home-state bias across fund family sizes. We proxy for the resources available to managers with the TNA of their fund family by grouping fund families into quintiles.15 The estimated coefficient on the interaction of FamSizeQuin and PctMgrHmState is −43 bps and is significant at the 5% level. The results indicate that funds belonging to families in the smallest quintile overweight their managers’ home states by 228 bps (significant at the 1% level), compared to just 13 bps for the largest funds. In column 4, we create a measure of fund-level resources (FundSizeQuin) that is analogous to that of our family size measure using fund TNA, and interact it with PctMgrHmState. The estimated coefficient on the interaction of FundSizeQuin and PctMgrHmState is negative, but insignificant. As shown in Table 1, many funds are team-managed. Additionally, in Table 2, we showed that the estimate on PctMgrHmState was higher for team-managed funds than for single-managed funds. However, previously we did not test this formally. In column 5 of the table, we investigate whether team-managed funds have greater cumulative home-state biases. Finding that they do would suggest that each manager in the management team gets his stock picks and that these picks are influenced by the manager’s home state. The coefficient on the interaction variable of interest is 62 bps, which suggests that team-managed funds have a significantly higher cumulative abnormal weight allocated to manager home states. Column 6 of the table shows that the previous results that funds with fewer family resources and that are team-managed have greater home-state biases hold when all of the fund characteristics explored in the table are included in the same regression. 2.3 Manager characteristics Additional insights can be gained by testing which types of managers are more prone to overweighting. In Table 5, we investigate whether differences in manager age, experience, home-state tenure, college location, and college quality are associated with the degree to which managers overweight their home states. In order to do so, we must expand the current regression format from quarterly fund-state observations to quarterly manager-fundstate observations. Using such a specification, we can include interactions

15 Quintiles are based on total fund TNA each quarter using the universe of funds covered by the CRSP Mutual

Fund Database and creating the variable FamSizeQuin, which is equal to the family TNA quintile ranking, minus one. We define the variable in this way so that the coefficient on PctMgrHmState has the interpretation of the home-state overweighting by funds in the smallest family size quintile.

2577

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2577

2563–2601

[10:29 2/7/2012 OEP-hhs075online.tex]

0.86 2,079,627 fund-state

(0.02)

(0.02) 0.86 2,226,507 fund-state

0.90∗∗∗

0.19∗∗∗ (0.06) −0.04 (0.08)

0.90∗∗∗

0.15∗∗∗ (0.05)

(2)

0.86 2,226,507 fund-state

(0.02)

0.90∗∗∗

−0.20∗∗ (0.08)

0.25∗∗∗ (0.07)

(3)

0.85 1,056,567 fund-state

(0.03)

0.90∗∗∗

0.41∗ (0.22)

−0.08 (0.10)

(4)

0.85 1,433,457 fund-state

(0.03)

0.89∗∗∗

−0.07 (0.06) 0.23∗∗ (0.11)

0.02 (0.05)

(5)

0.85 1,433,457 fund-state

(0.03)

0.89∗∗∗

0.01 (0.05)

0.14∗∗∗ (0.04)

(6)

0.86 1,836,102 fund-state

−0.08 (0.09) 0.90∗∗∗ (0.03)

0.13 (0.12)

(7)

0.85 1,056,567 fund-state

0.90∗∗∗ (0.03)

−0.36∗∗ (0.15) 0.39∗ (0.22)

0.14∗ (0.07)

(8)

0.85 1,433,457 fund-state

0.89∗∗∗ (0.03)

−0.07 (0.06) 0.24∗∗ (0.11)

−0.24∗∗ (0.10)

(9)

where wi,s,t is the portfolio weight fund i allocates to firms headquartered in state s during quarter t , MgrHmStateDumi,j,s,t is a dummy variable that is one if state s is the home state of manager j of fund i during quarter t , MorningstarBMWti,s,t is the average portfolio weight in state s of all funds within the same Morningstar category as fund i during quarter t , λi,s are fund-state fixed effects, and Controlsi,j,s,t is a vector of relevant control variables. The sample includes 2,226,507 quarterly manager-fund-state observations from the first quarter of 1996 to the fourth quarter of 2009 and includes 1,767 unique funds managed by 2,109 unique managers. The regressions include interaction terms with various fund manager characteristics. Observations are quarterly fund-manager-state observations. AgeGTMed is a dummy variable that is one if the age of the manager is greater than the median in the sample. ExperienceGTMed is a dummy variable that is one if the mutual fund managing experience of the manager is greater than the median in the sample. HmTenureGTMed is a dummy variable that is one if the number of years that the manager lived in his home state is greater than the median in the sample (see the appendix for a description of how this is calculated). MgrCollState is a dummy variable that is one if the manager went to college in the state. IvyLeague is a dummy variable that is one if the manager has a degree from an Ivy League school. All specifications include a constant and the main effect for the interaction variables being tested (AgeGTMed, ExperienceGTMed, HMTenureGTMed, and IvyLeague), although for brevity we do not report these coefficient estimates. Standard errors, clustered at the fund level, are in parentheses. Significance levels are denoted by *, **, and ***, which correspond to the 10%, 5%, and 1% levels, respectively.

wi,s,t = λi,s +β MgrHmStateDumi,j,s,t +δ MorningstarBMWti,s,t +  Controlsi,j,s,t +i,j,s,t ,

The table reports the coefficient estimates and standard errors from the OLS estimation of various forms of the regression equation

AdjR2 N Fixed effects

MorningstarBMWt

MgrHmStateDum × IvyLeague

MgrHmStateDum × MgrCollState

MgrCollState

MgrHmStateDum × HmTenureGTMed

MgrHmStateDum × ExperienceGTMed

MgrHmStateDum × AgeGTMed

MgrHmStateDum

(1)

Table 5 Manager characteristics and home-state overweighting

The Review of Financial Studies / v 25 n 8 2012

2578

Page: 2578

2563–2601

No Place Like Home

with manager-specific characteristics that we would otherwise not be able to include in the fund-state quarterly observation framework. This specification also allows us to conduct more powerful within-fund tests that rely not only on manager turnover within funds, but also on variation in the home-state origin of managers of the same fund during the same quarter. Specifically, we estimate wi,s,t = λi,s +βMgrHmStateDumi,j,s,t +δMorningstarBMWti,s,t +  Controlsi,j,s,t +i,j,s,t ,

(2)

which differs from regression equation (1) in the variable of interest, MgrHmStateDumi,j,s,t , which is a dummy variable that is one if manager j of fund i is from state s and is zero otherwise; in the control and interaction variables that can now also depend on manager j ; and in the inclusion of fund-state fixed effects, λi,s . In column 1, we estimate regression equation 2, including only the Morningstar benchmark weights and fund-state fixed effects. The estimate on MgrHmStateDum is 15 bps and statistically significant at the 1% level. It is important to note that the estimate on MgrHmStateDum is on a per-manager basis and is biased downward from the estimate in column 8 of Table 2. This is because the home-state bias is smaller per manager in team-managed funds, and in this specification, team-managed funds have more weight. The cummulative home-state bias for a given fund is the sum of its managers’ biases.16 We test for differences in manager home-state overweighting by interacting various dummy variables with MgrHmStateDum using this conservative within-fund estimation specification.17 If information is driving the portfolio overweighting, then we would expect to find that the observed portfolio overweighting is more common among more experienced managers. If, however, a bias is driving the result, we should find that inexperienced managers overweight their home states more. We test this hypothesis in columns 2 and 3 of the table by interacting MgrHmStateDum with dummy variables indicating managers who are older than the median age in the sample (AgeGTMed) and managers who have greater than the sample median mutual fund management experience (ExperienceGTMed). The regression estimates show that although manager age does not affect the manager’s home-state bias, the portfolio management experience of the manager does. In fact, only managers who are early in their careers exhibit the home bias in their portfolio holdings: the estimate on ExperienceGTMed is −20 bps and is significant at the 5% level. 16 Since the average fund has on average 2.5 managers, this implies that the total home-state overweighting of

the average fund should be approximately 38 bps, which is about 10 bps lower than specification 8 of Table 2 because of the previously mentioned downward bias. 17 All regressions include the level of each interaction variable, although for brevity we supress coefficient estimates.

2579

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2579

2563–2601

The Review of Financial Studies / v 25 n 8 2012

Both the information and familiarity hypotheses suggest that the longer a manager lives in his home state the more likely he is to overweight his state: he could have more contacts at local firms, or could simply be more familiar with the companies of the state. To capture this, we first estimate how long managers lived in their home states and then construct a dummy, HmTenureGTMed, that equals one when home tenure is greater than the median in the sample (see the appendix). The coefficient estimate on HmTenureGTMed is significantly estimated at 41 bps, which indicates that the overweighting of home-state stocks is concentrated among managers who spent more time in their home states. Another way to examine home-state tenure is to consider whether the manager attended college in-state. Consistent with the home tenure results, we would expect managers who remain in their home state for college to be even more familiar with companies in that state. We test this by interacting a dummy variable that equals one if the manager attended college in state s (MgrCollState) with MgrHmStateDum. The results in column 4 confirm that home-state overweighting is strongest for managers who attended college instate. In column 5, we confirm that attending college in a state without also having grown up there has no effect on overweighting. Finally, we look at whether more educated managers have lower home-state biases. If information drives the home-state bias, then more-educated managers should overweight home-state stocks more. If, however, the home-state bias is driven by familiarity, we might expect less-educated managers to exhibit stronger biases. To test this, we interact MgrHmStateDum with IvyLeague, which is a dummy variable that equals one if the manager earned a degree from an Ivy League institution. The estimate on the interaction is negative, but not statistically different from zero. There is no evidence that more-educated managers avoid or seek out home-state securities.18 The regressions in columns 8 and 9 show that when both experience and measures of home-state tenure are included in the same regression, the results that more experienced managers have significantly lower home-state biases and managers who live in their home states longer have significantly larger home-state biases continue to hold. 2.4 Stock characteristics We now turn to the question of which types of stocks managers overweight in their home states. It would be surprising if the home bias were driven by fund holdings in well-known companies. If familiarity affects holdings, then all fund managers can be expected to hold well-known companies, regardless of where their home state is. This would increase the benchmark weight in a stock, thereby raising the bar for other managers to overweight the stock. Moreover, a manager is less likely to believe he has unique information about a very well-known stock from his home state.

18 We found similar results using school rankings and average SAT scores.

2580

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2580

2563–2601

No Place Like Home

We test how stock characteristics are related to manager home-state overweighting by estimating regressions similar to that in column 3 of Table 2 using quarterly fund-stock observations for subsamples split by stock characteristics. We use sample splits instead of interacting the variables for ease of interpretation. Most of the characteristics on which we split the sample are correlated with firm size. Larger firms tend to have larger average portfolio weights, so the estimate on the interaction term tells us whether the home-state overweighting is different on an absolute level, but not on a relative (to the average portfolio weight) level. We estimate wi,k,t = βPctMgrHmStatei,k,t + δMorningstarBMWti,k,t +  Controlsi,k,t + i,k,t ,

(3)

where wi,k,t is the portfolio weight fund i allocates to security k during quarter t. The regression includes all stocks within a fund’s investment universe, regardless of whether the fund holds the stock, so many of the observations have zero weight. Results are reported in Table 6. In column 1, we summarize the results for the full sample. Compared with the average stock weight of 5.5 bps, the excess holding in home states is 0.9 bps, representing an overweighting of 16.3%, which is similar to our previous estimates. In columns 2 and 3, we report results for stocks that are, and are not, included in the S&P 500 index, respectively. Overweighting of home-state stocks occurs in both subsamples, but relative to benchmark weights, stocks that are not in the index are overweighted at about twice the rate of stocks that are in the index (23.0% vs. 10.4%). We see similar results when splitting the sample along several other dimensions: sales, analyst coverage, and advertising. For each variable, we divide the sample into stocks that are above and below the median value of the characteristic each quarter. In each case, overweighting in home-state stocks is higher in both subsamples. When compared with the benchmark weights, overweighting is higher for stocks with low levels of sales (20.8% vs. 14.9%), low levels of analyst coverage (26.2% vs. 13.1%), and low levels of advertising (26.4% vs. 16.2%). Note as well that in each case it is more difficult to explain the holdings of small, less-known stocks than the more prominent stocks, as indicated by the comparatively low R 2 s. 3. Do Funds Outperform in Their Managers’ Home-state Stocks? 3.1 Portfolio performance Our results thus far show that fund managers invest significantly more in stocks that are headquartered in their home states than do those who manage peer funds but grew up elsewhere. Moreover, the previous section suggests that

2581

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2581

2563–2601

[10:29 2/7/2012 OEP-hhs075online.tex]

0.0553 16.27

Mean stock weight Percent home-state overweight 0.1468 10.42

0.19 10,007 0.0318 22.97

0.05 38,837

0.0073∗∗∗ (0.0015) 0.0084∗∗∗ (0.0013) 1.0205∗∗∗ (0.0167) −0.0010∗∗ (0.0005)

Non-S&P 500 stocks (3)

0.0857 14.93

0.16 23,749

0.0128∗∗∗ (0.0032) 0.0177∗∗∗ (0.0026) 0.9901∗∗∗ (0.0127) −0.0010 (0.0010)

High sales (4)

0.0265 20.76

0.05 23,747

0.0055∗∗∗ (0.0015) 0.0075∗∗∗ (0.0014) 1.0360∗∗∗ (0.0216) −0.0012∗∗ (0.0005)

Low sales (5)

0.0817 13.10

0.16 24,937

0.0107∗∗∗ (0.0028) 0.0155∗∗∗ (0.0024) 0.9908∗∗∗ (0.0125) −0.0008 (0.0009)

High analyst cov. (6)

0.0279 26.16

0.05 23,977

0.0073∗∗∗ (0.0017) 0.0078∗∗∗ (0.0013) 1.0134∗∗∗ (0.0174) −0.0008∗ (0.0005)

Low analyst cov. (7)

0.0987 16.22

0.19 8,180

0.0160∗∗∗ (0.0049) 0.0199∗∗∗ (0.0041) 0.9889∗∗∗ (0.0137) −0.0011 (0.0011)

High advertising (8)

0.0284 26.37

0.06 8,179

0.0075∗∗∗ (0.0020) 0.0091∗∗∗ (0.0018) 1.0174∗∗∗ (0.0173) −0.0010∗∗ (0.0004)

Low advertising (9)

where wi,k,t is the portfolio weight fund i allocates to security k during quarter t , PctMgrHmStatei,k,t is the percentage of fund i ’s managers during quarter t whose home state is the same as security k ’s state of headquarters, MorningstarBMWti,k,t is the average portfolio weight in security k of all funds within the same Morningstar category as fund i during quarter t , and Controlsi,k,t is a vector of control variables. The sample includes 48,914,091 quarterly fund-stock observations from the first quarter of 1996 to the fourth quarter of 2009. Stocks held by at least one fund in the same nine-box Morningstar category are included for each fund during the quarter. MFHQStatei,k,t is a dummy variable that takes a value of one if the mutual fund complex of fund i and stock k are headquartered in the same state during quarter t and is zero otherwise. In column 1, the regression is estimated for the full sample. In columns 2 and 3, stocks included and excluded from the S&P 500 index are included in the samples, respectively. In columns 4 through 9, regressions are estimated for samples split by the median level of sales, analyst coverage, and advertising, respectively. Standard errors, clustered at the fund level, are in parentheses. Significance levels are denoted by *, **, and ***, which correspond to the 10%, 5%, and 1% levels, respectively. Reported at the bottom of each column is the mean weight on each stock in the sample and the percent home-state overweighting, which is the coefficient estimate on PctMgrHmState divided by the mean stock weight in the sample.

wi,k,t = β PctMgrHmStatei,k,t +δ MorningstarBMWti,k,t +  Controlsi,k,t +i,k,t ,

The table reports the coefficient estimates and standard errors from the OLS estimation of the regression equation

0.14 48,914

0.0153∗∗∗ (0.0058) 0.0243∗∗∗ (0.0045) 0.9887∗∗∗ (0.0142) −0.0019 (0.0017)

0.0090∗∗∗ (0.0019) 0.0118∗∗∗ (0.0016) 0.9929∗∗∗ (0.0114) −0.0006 (0.0006)

AdjR2 N (thousands)

Intercept

MorningstarBMWt

MFHQState

PctMgrHmState

S&P 500 stocks (2)

All stocks (1)

Table 6 Stock characteristics and home-state overweighting

The Review of Financial Studies / v 25 n 8 2012

2582

Page: 2582

2563–2601

No Place Like Home

this overweighting may be due to familiarity. In this section, we formally test the information and familiarity hypotheses using direct, performance-based analyses. The two hypotheses offer very different empirical predictions concerning performance. If managers overweight securities headquartered in their home states due to a comparative advantage they have in generating information about these companies, the home-state portion of their portfolios should outperform their other holdings. In contrast, the performance effect of familiarity should be nonpositive and depend on whether managers have any skill in general. If managers do not generally possess a superior ability to pick stocks, choosing based on familiarity will produce outcomes that are no worse than any of their other stock selection methods. In this case, familiarity will have no effect on performance. Alternatively, if managers are skilled, familiarity should lead to a lower return on home-state investments since the choice of these stocks, unlike those of other stocks in the manager’s portfolio, is motivated by behavioral rather than informational reasons. Whether fund managers have skill is an ongoing debate in the literature. Many studies find that active mutual funds as a whole underperform their passive benchmarks (Jensen 1968; Gruber 1996; Chevalier and Ellison 1999a; Carhart 1997). This may not mean that fund managers are unskilled, however. Several recent studies argue that not all fund holdings are expected to outperform; instead, many of the fund’s trades may simply reflect the incentives managers face in the mutual fund industry. For example, managers often make investment choices that allow them to stay close to their peers or appear attractive on marketing materials that target specific clienteles. These studies argue that analyzing the performance of the total portfolio can be misleading and suggest looking for evidence of stock selection ability instead in a subset of holdings that is most likely to be informed (Alexander, Cici, and Gibson 2007; Cremers and Petajisto 2009; Pomorski 2009; Cohen, Polk, and Silli 2010). In our context, the findings in Coval and Moskowitz (2001) suggest that local stocks would be one such subset. We use this insight to test our information and familiarity hypotheses. Since the familiarity hypothesis predicts that the fund’s home-state stocks will underperform the manager’s informed choices, but will do no better or worse than the rest of his portfolio, we group the holdings of each fund into three disjoint categories. In particular, we first create a portfolio of home-state stocks (“home”), and divide the remaining securities into two groups: those stocks that are most likely to represent managers’ informed picks, and the rest. Relying on the results in Coval and Moskowitz (2001), we use local stocks for the former. We define a stock as local when it is headquartered in the state where the fund is located.19 Therefore, this further classification creates a “local” and an “away” portfolio.

19 When a stock is both local and a home-state stock, we classify it as local.

2583

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2583

2563–2601

The Review of Financial Studies / v 25 n 8 2012

To compare the performance of the home-state holdings to those of the local and away holdings separately, we calculate home, local, and away portfolio returns for each fund manager-quarter:    wi,k,t H  Ri,t = rk,t+1 , (4) k∈H\L wi,k,t k∈H\L L Ri,t

=

 k∈L

and A Ri,t =

 wi,k,t  rk,t+1 , k∈L wi,k,t

  k ∈H∪L /



(5)

 wi,k,t rk,t+1 , wi,k,t k ∈H∪L /

(6)

where H is the set of firms headquartered in a manager’s home state, L is the set of firms headquartered in the state where the fund is located, and wi,k,t is the actual portfolio weight of fund i in stock k during quarter t. We then create home, local, and away portfolio returns by calculating the weighted average of the returns in equations (4)–(6) across funds at time t, weighting each fund’s return by its TNA. We use two different excess return measures for rk : the average monthly raw return in excess of the one-month Treasury bill rate (obtained from Kenneth French’s website), and the average monthly benchmark-adjusted excess return as in Daniel, Grinblatt, Titman, and Wermers (1997, “DGTW”) and Wermers (2005).20 The DGTW benchmark adjustment procedure sorts stocks into size quintiles, and within each size quintile stocks are further sorted into bookto-market quintiles. Finally, each book-to-market quintile is divided into five momentum portfolios. The sorting process creates 125 stock characteristics groups, for which benchmark portfolios are formed by calculating the valueweighted average return of the stocks in each category. We calculate the DGTW benchmark returns in two ways: first, we include the CRSP universe of common stocks in the calculation; second, we limit our sample to stocks with prices above $5, as funds often face restrictions on investments in low-priced stocks. Finally, with each of the DGTW benchmark portfolio returns, we calculate the characteristic-adjusted returns by subtracting from each stock’s return the value-weighted average return of stocks with similar size, book-to-market, and momentum characteristics as defined by the triple-sort benchmark portfolios. The average returns for the home (R H ), local (R L ), and away (R A ) portfolios over the 56 quarters and the difference between these averages are presented in Table 7. Column 2 of the table shows that the home portfolio earns, on

20 The DGTW benchmarks are available at http://www.smith.umd.edu/faculty/rwermers/ftpsite/Dgtw/coverpage.

htm.

2584

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2584

2563–2601

No Place Like Home

Table 7 Home-state portfolio performance RH

RL

RA

R HNH

R H −R L

R H −R A

R H −R HNH

Excess return

0.44 (0.30)

0.70 (0.12)

0.44 (0.31)

0.42 (0.40)

−0.26 (0.67)

0.01 (0.99)

0.02 (0.97)

DGTW (entire CRSP universe)

0.08 (0.36)

0.33∗∗∗ 0.06 (0.00) (0.22)

0.10 (0.19)

−0.25∗ (0.05)

0.02 (0.84)

−0.03 (0.82)

DGTW (CRSP with price>$5)

0.00 (0.98)

0.30∗∗∗ −0.02 (0.00) (0.67)

0.05 (0.57)

−0.30∗∗ (0.03)

0.02 (0.85)

−0.05 (0.67)

The table reports returns for home-state (H ), local-state (L), and away (A) holdings of fund managers, as well as the returns of the portfolio of home-state stocks that are not held (H NH ). We use the performance measures H= Ri,t

L = Ri,t



 k∈H\L



wi,k,t k∈H\L wi,k,t



wi,k,t k∈L wi,k,t



k∈L

 rk,t+1 ,

 rk,t+1 ,

and A = Ri,t

  k ∈H∪L /



wi,k,t wi,k,t k ∈H∪L /

 rk,t+1

to calculate the home, local, and away returns each quarter by averaging across funds in quarter t , weighting each fund’s return by its total net asset value. (H denotes the set of firms headquartered in a manager’s home state, L is the set of firms headquartered in the state where the fund is located, and wi,k,t is the actual portfolio weight of fund i in stock k during quarter t .) The portfolios of not held stocks are value-weighted. We then average these portfolios across funds in quarter t , weighting each fund’s “not-held” return by its total net asset value. Columns 2–5 report the raw and risk-adjusted returns for the home, local, away, and not-held portfolios. Risk adjustment is based on Daniel, Grinblatt, Titman, and Wermers (1997, “DGTW”). Columns 6–8 report the pairwise difference in returns between the home portfolio and each of the other portfolios, and the results of the paired t-test testing whether the difference is significantly different from zero. Significance levels for the paired t-tests are denoted by *, **, and ***, which correspond to the 10%, 5%, and 1% levels, respectively. p-values are reported in parentheses.

average, 44 bps per month in excess of the risk-free rate (i.e., 5.28% per year). It outperforms its size, BE/ME, and momentum benchmark portfolio by an average of 8 bps per month (0.96% per year), but this characteristic-adjusted performance is not statistically significant. Moreover, when we construct our DGTW benchmarks from stocks with a price above $5, the home portfolio underperforms its characteristics-based benchmark by −0.2 bps per month. In contrast, the local portfolio exhibits much higher raw and risk-adjusted performance. The risk-adjusted returns of the local portfolio range between 30 and 33 bps per month (3.6–4.0% per year), both significant at better than the 1% level. This result confirms that local holdings may indeed represent the manager’s information-motivated trades and is consistent with the results reported in Coval and Moskowitz (2001).21 Finally, column 4 reports that the

21 Sulaeman (2008) shows that local stocks do not significantly outperform distant holdings during 1994–2004. In

unreported analyses, we confirm that our sample is consistent with this finding by estimating the performance of the local portfolio during the first part of the sample period.

2585

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2585

2563–2601

The Review of Financial Studies / v 25 n 8 2012

raw and risk-adjusted performance of the away portfolio is very similar to that of the home portfolio and is not significantly different from zero. Columns 6 and 7 formally test the difference between the performance of the home and local, and the home and away, portfolios, respectively. On a risk-adjusted basis, the local portfolio outperforms the home portfolio by 25–30 bps per month (3.0–3.6% per year), and both of these benchmark-adjusted return differences are statistically significant. In contrast, there is no statistically significant difference between the performance of the home and the away portfolios. These results suggest that managers do not appear to have a comparative advantage in choosing which home-state stocks to buy. It is possible, however, that they do better at selecting which home-state stocks to avoid. Since mutual funds are generally restricted from short-selling, any negative information that managers have about their home securities should manifest itself in the performance of stocks not held by the fund. To investigate this, we create a value-weighted portfolio of home-state stocks that are not held by each fund. We then average the returns of these portfolios across funds, weighting by each fund’s TNA. Column 5 of Table 7 reports the raw and risk-adjusted performance of home-state, but not held, stocks. Both the raw and the characteristics-adjusted values are very similar to those of held stocks described in column 2. Finally, column 8 reports a test of the difference between the returns of home-state held and not-held stocks. The difference between the two sets is very small and statistically insignificant. If anything, not-held stocks outperform held stocks by a few basis points, indicating that managers do not have adverse information about home securities either. The results in Table 7 provide strong evidence against the information hypothesis and support the familiarity bias explanation. A potential concern with our analysis, however, is that under the information hypothesis, only managers better able to acquire information about home companies concentrate their holdings in their home state. Therefore, it is possible that our approach of averaging the home-state performance across all managers at time t masks the real performance effect of the portfolio tilt. To examine this, we sort funds into deciles based on the magnitude of their home-state bias. We then determine the quarterly home, local, and away returns for each decile by averaging the performance measures in (4)–(6), respectively, across the decile funds at time t, weighting each fund’s return by its TNA. We define home-state bias as the difference between the fund’s actual weight in the manager’s home state and the weight the average Morningstar benchmark fund places on this state during the same quarter. In untabulated analyses, we compare home and local, and home and away, portfolio returns, respectively, within each decile. We find that local holdings outperform the home portfolio in every home bias decile, though the difference is not always statistically significant. In contrast, the home and away portfolios outperform each other with roughly the same frequencies,

2586

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2586

2563–2601

No Place Like Home

and their performance difference is not significant in any of the bias decile groups.22 Finally, in Table 6, we report that funds overweight home-state stocks more— on a relative basis—when they are not included in the S&P 500, are smaller, or have fewer analysts, for example. In unreported analyses, we examine the performance of subportfolios of home-state investments based on these stock characteristics. We find that the returns of the home-state subportfolios are generally not significantly different from zero. This is also consistent with the perceived information story whereby managers erroneously believe they have an information advantage for less-known securities. 3.2 Fund diversification Our performance results suggest that, in the absence of informed investment ideas, using familiarity to build a portfolio will not harm the fund’s returns. So, does overweighting home-state stocks matter at all? That is, is there a cost to overweighting itself? We next investigate underdiversification as one possibility, and study whether funds whose managers exhibit larger home-state biases have greater idiosyncratic volatility. We measure idiosyncratic volatility as the quarterly average of the assetweighted average idiosyncratic volatility of mutual funds within each homestate bias decile for the 46 quarters between the second quarter of 1998 and the fourth quarter of 2009.23 Idiosyncratic volatility for each fund is computed using daily returns from the quarter immediately following the measurement of the home-state bias. In Table 8, we report the average idiosyncratic volatility by home-statebias decile for our sample of mutual funds. Column 3 reports the average idiosyncratic volatility for each decile calculated from the residuals of the CAPM, and column 6 uses the residuals from the four-factor model. Interestingly, funds with the greatest home-state bias also have the highest average idiosyncratic volatility. It is between 22% and 41% larger than that of funds with little to no bias (deciles 4, 5, and 6).24 Columns 4 and 7 show the pairwise difference in the average idiosyncratic volatility between decile 10 and each of the other deciles, and the results of the paired t-test testing whether the difference is significantly different from zero.

22 In addition to the tests described in the text, we perform multivariate performance analyses that control for the

location of the management company and incorporate penalties for putting a larger weight on bad performers or a smaller weight on good performers relative to peer funds. In all tests, we find no evidence supporting the information hypothesis. 23 Equal-weighted idiosyncratic volatility yields similar results. The sample period differs slightly due to availability

of daily return data from the CRSP mutual fund database. 24 The “U”-shape relationship between idiosyncratic volatility and the home-state bias can be attributed to the fact

that if a fund manager is severely underweighting firms in her home state, she must be overweighting firms from other states.

2587

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2587

2563–2601

[10:29 2/7/2012 OEP-hhs075online.tex]

1 2 3 4 5 6 7 8 9 10

Idio. vol. (%) 0.41 0.37 0.37 0.34 0.36 0.33 0.35 0.37 0.39 0.44

Avg. homestate bias (%) −5.16 −2.00 −1.09 −0.52 −0.12 0.33 1.00 1.97 3.58 10.50

% of qtrs. < decile 10 65.2 87.0 82.6 89.1 87.0 93.5 89.1 91.3 82.6

CAPM Diff. from decile 10 (%) 0.03∗∗ 0.07∗∗∗ 0.07∗∗∗ 0.09∗∗∗ 0.08∗∗∗ 0.11∗∗∗ 0.09∗∗∗ 0.07∗∗∗ 0.05∗∗∗ 0.32 0.28 0.28 0.26 0.26 0.24 0.26 0.28 0.30 0.35

Idio. vol. (%)

0.03∗∗∗ 0.07∗∗∗ 0.06∗∗∗ 0.09∗∗∗ 0.08∗∗∗ 0.10∗∗∗ 0.08∗∗∗ 0.06∗∗∗ 0.05∗∗∗

Four-factor model Diff. from decile 10 (%)

67.4 91.3 91.3 95.7 95.7 95.7 95.7 93.5 87.0

% of qtrs. < decile 10

The table reports idiosyncratic volatility for deciles of the home-state bias. Home-state bias is defined as the fund’s actual weight in the manager’s home state s minus the portfolio weight of the fund’s Morningstar benchmark in state s during the same time period. For funds with multiple managers, the average home-state bias of the fund’s managers is used to create the decile portfolios. Idiosyncratic volatility is measured as the quarterly average of the asset-weighted average idiosyncratic volatility of mutual funds within each home-state bias decile for the 46 quarters between the second quarter of 1998 and the fourth quarter of 2009. Idiosyncratic volatility for each fund is computed using daily gross returns from the quarter following the measurement of the home-state bias. Column 3 reports the average idiosyncratic volatility for each decile calculated from the residuals of the CAPM, and column 6 uses the residuals from the four-factor model. Columns 4 and 7 show the differences in the average idiosyncratic volatility from decile 10, and stars indicate the significance level of the paired t-test testing if the difference is signficantly different from zero. Significance levels are denoted by *, **, and ***, which correspond to the 10%, 5%, and 1% levels, respectively. Columns 5 and 8 display the percentage of quarters that the average idiosyncratic volatility of funds in the corresponding decile is less than the average idiosyncratic volatility of the funds in decile 10.

high

low

Home-state bias decile

Table 8 Mutual fund idiosyncratic volatility by home-state bias deciles

The Review of Financial Studies / v 25 n 8 2012

2588

Page: 2588

2563–2601

No Place Like Home

The tests show that idiosyncratic volatility of mutual funds in the highest homestate bias decile is significantly larger than in all other deciles. Standard portfolio theory says that as more stocks are added to a portfolio, the idiosyncratic volatility of that portfolio should fall. It is possible that funds with the largest home-state biases have fewer holdings than other funds, and that this leads mechanically to the results in Table 8. Alternatively, higher idiosyncratic volatility could come from a strong correlation between homestate securities, which is consistent with the results of Pirinsky and Wang (2006), who show that the returns of stocks headquartered in the same area exhibit substantial comovement. The effect is especially strong among smaller stocks, which is precisely where home-state overweighting is strongest. According to their results, concentrating portfolio holdings in one particular area versus diversifying geographically should lead to higher idiosyncratic volatility. In order to isolate the effect of home-state overweighting on idiosyncratic volatility, we regress daily idiosyncratic volatility on a dummy variable that indicates whether the fund is in the highest home-state bias decile at the end of the previous quarter (HmStateBiasDecile10) and lagged fund characteristics.As in Table 8, idiosyncratic volatility for each fund is computed from the residuals of the four-factor model using daily gross returns from the quarter following the measurement of the home-state bias. The regression results are reported in Table 9. In column 1, the coefficient estimate on HmStateBiasDecile10 is 5.7 bps. The average daily idiosyncratic volatility for funds in the sample is 33.4 bps. This suggests that funds that overweight their managers’ home states the most have approximately 17% higher idiosyncratic volatility than other funds in the sample. This result is similar to the magnitude found in the previous table. In column 2, we control for the number of stocks in the portfolio by including the natural log of the number of stocks held during the previous quarter (LogNumStocksHeld). As expected, the coefficient estimate on LogNumStocksHeld is negative and explains a substantial amount of the variation in idiosyncratic volatility across funds. Including this control also reduces the coefficient estimate on HmStateBiasDecile10 to 2.69 bps, but the estimate remains significant at better than the 1% level. In columns 3 and 4, we include additional control variables to control for fund size, whether the fund is team-managed, and manager experience. The inclusion of these controls reduces the magnitude of the coefficient estimate on HmStateBiasDecile10 slightly, but it remains statistically significant. In columns 5 through 7, we repeat the specifications in columns 2 through 4 with the inclusion of quarter fixed effects. The explanatory power of these models increases substantially. While the inclusion of these fixed effects wipes out the significance of team-managed funds and manager experience, they do not significantly alter the coefficient estimates on HmStateBiasDecile10. In the most conservative fixed effects specification, the coefficient estimate on HmStateBiasDecile10 is 2.65 bps, suggesting that beyond the effects of other fund characteristics, funds with the largest home-state biases have nearly 2589

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2589

2563–2601

[10:29 2/7/2012 OEP-hhs075online.tex]

0.3343

No

0.09 22,343

0.6530∗∗∗ (0.0391)

0.3343

No

0.09 22,343

0.7808∗∗∗ (0.0504)

0.0234∗∗ (0.0101) −0.0683∗∗∗ (0.0058) −0.0076∗∗∗ (0.0020) −0.0201∗∗ (0.0083)

(3)

0.3343

No

0.10 22,343

0.0235∗∗ (0.0101) −0.0707∗∗∗ (0.0061) −0.0063∗∗∗ (0.0021) −0.0225∗∗∗ (0.0086) −0.0023∗∗ (0.0010) 0.7864∗∗∗ (0.0502)

(4)

0.3343

Quarter

0.46 22,343

0.7000∗∗∗ (0.0235)

0.0276∗∗∗ (0.0090) −0.0735∗∗∗ (0.0055)

(5)

0.3343

Quarter

0.47 22,343

0.8355∗∗∗ (0.0370)

0.0265∗∗∗ (0.0090) −0.0660∗∗∗ (0.0053) −0.0088∗∗∗ (0.0016) 0.0045 (0.0070)

(6)

0.3343

Quarter

0.47 22,343

0.0265∗∗∗ (0.0090) −0.0657∗∗∗ (0.0053) −0.0089∗∗∗ (0.0016) 0.0049 (0.0070) 0.0003 (0.0006) 0.8350∗∗∗ (0.0370)

(7)

The table reports the coefficient estimates and standard errors from the OLS estimation of fund idiosyncratic volatility on a dummy variable that indicates whether the fund is in the highest home-state bias decile at the end of the previous quarter (HmStateBiasDecile10) and lagged fund characteristics. Idiosyncratic volatility for each fund is computed using daily gross returns from the quarter following the measurement of the home-state bias and is calculated from the residuals of the four-factor model. All funds from the original sample are included for the 46 quarters between the second quarter of 1998 and the fourth quarter of 2009. The home-state bias is defined as the fund’s actual weight in the manager’s home state s minus the portfolio weight of the fund’s Morningstar benchmark in state s during the same time period. For funds with multiple managers, the average home-state bias of the fund’s managers is used. Included as control variables are the natural logarithm of the number of stocks held by the fund LogNumStocksHeld, the natural logarithm of the fund size, measured by TNA (LogFundSize), a dummy variable that is one if the fund has more than one manager (TeamManaged), and the average number of years of experience of the management team (MgrExperience). Regressions in columns 5 through 7 include quarter fixed effects. Standard errors, clustered by fund and quarter, are in parentheses. Significance levels are denoted by *, **, and ***, which correspond to the 10%, 5%, and 1% levels, respectively.

0.3343

No

Fixed Effects

Mean idiosyncratic volatility

0.01 22,343

0.3287∗∗∗ (0.0185)

0.0269∗∗∗ (0.0098) −0.0754∗∗∗ (0.0058)

0.0565∗∗∗ (0.0105)

AdjR2 N

Intercept

MgrExperience

TeamManaged

LogFundSize

LogNumStocksHeld

HmStateBiasDecile10

(2)

(1)

Table 9 Home-state oveweighting and fund idiosyncratic volatility

The Review of Financial Studies / v 25 n 8 2012

2590

Page: 2590

2563–2601

No Place Like Home

8% higher idiosyncratic volatility than other funds.25 These findings suggest that mutual funds whose managers exhibit the largest familiarity biases hold inefficient portfolios from exposing their investors to excess risk. 4. Robustness 4.1 Subsample analysis In Table W2 of the Online Appendix, we perform a number of robustness checks using our baseline regression from column 3 in Panel B of Table 2. In column 1 of Table W2, we add MFAdvStatei,s,t , which is a dummy variable that takes a value of one if the adviser of fund i is headquartered in state s during quarter t and is zero otherwise. The coefficient estimate of PctMgrHmState is 65 bps and is highly statistically significant after the inclusion of MFAdvState. In column 2, we reestimate our model including only observations where fund i during quarter t does not employ a subadviser. Our results remain robust in this subsample. We showed in Figure 1 that many funds are headquartered in New York. In column 3 of the table, we estimate our baseline regression using only those observations where the fund is headquartered outside New York. The coefficient estimate on PctMgrHmState is 95 bps and is statistically significant at greater than the 1% level. We estimate a similar regression in column 4, omitting all observations where the fund is headquartered in any of the financial centers of California, Illinois, New York, Massachusetts, or Pennsylvania. Again, the estimate on PctMgrHmState is positive and significant. In columns 5 and 6, we omit observations of states that are less than 500 and 1,000 miles from the mutual fund complex headquarters, respectively. In both cases, the coefficient estimate on PctMgrHmState remains positive and significant at better than the 1% level. Finally, in column 7, we restrict the coefficient estimate on Morningstar BMWt to be equal to one. The coefficient estimate on PctMgrHmState remains positive and significantly estimated at 81 bps. 4.2 Portfolio distance An alternative to investigating whether portfolio managers overweight their holdings of firms headquartered in their home states is to test whether the stocks they hold are geographically closer to their home states than those of a benchmark portfolio. Since we know the state where the manager grew up, we need to choose a location in that state that represents the “middle” of the state. Rather than using the geographic centroid of each state, we use a populationweighted measure, which takes into account where people live. In other words, we choose the point in each state that minimizes the expected distance to a randomly selected person who lives in that state. These population-weighted 25 The results are similar when using the CAPM to compute idiosyncratic volatility (untabulated). Additionally,

using the home-state bias in place of HmStateBiasDecile10 yields similar results.

2591

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2591

2563–2601

The Review of Financial Studies / v 25 n 8 2012

Figure 2 Excess portfolio weights by distance from managers’ home states The figure shows the average excess weight a fund has in a stock as a function of the distance between the stock issuer’s headquarters and fund manager’s home state. Observations are at the fund-quarter-manager-stock level. For stock issuers’ locations, we use the centroid of the zip code in which the firm headquarters is located. For manager home-state locations, we use the population-weighted center of the state based on data from the 2000 census (see Figure W1 of the Online Appendix for details). Portfolio weights are in excess of the equally weighted average for all funds in the same nine-box Morningstar category on each quarter-end. The 5% confidence interval, based on standard errors that are clustered at the fund level, is shown with shading. The average weight in a security is 5.5 bps.

centers are calculated by the U.S. Census Bureau using data from the 2000 Census, and are shown in Figure W1 of the Online Appendix. Figure 2 plots the excess weight for stocks (relative to the Morningstar category benchmark portfolio) held by a manager as a function of the stock issuer’s distance from the fund manager’s home state. Specifically, for each stock held by a fund, we calculate the distance from the population-weighted center of the fund manager’s home state to the centroid of the zip code in which the stock issuer’s headquarters is located. We then compute the average excess weight for seven distance “bins.” Confidence intervals are based on clustered standard errors to allow for correlation within each fund. The average excess weight in stocks with headquarters located within 50 miles of a fund manager’s home state is 0.75 bps, and then declines for stocks located farther away. For comparison, the average stock weight is 5.5 bps, which implies an overweighting of 13.6%, and is comparable to the estimates reported earlier.26 26 Note that, in contrast to the results reported in Table 6, this analysis uses fund-manager-stock-quarter

observations. Following Coval and Moskowitz (1999), we also compute the value-weighted distance between

2592

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2592

2563–2601

No Place Like Home

4.3 Alternative geographic classifications In unreported analyses, we estimate the models from columns 2, 3, and 7 of Table 2 using two alternative geographic classifications instead of states: U.S. Census divisions and U.S. Census regions (see Figure W1 of the Online Appendix). Observations are quarterly fund-division/region observations. We create the variables PctMgrHmDivision and PctMgrHmRegion, which are defined analogously to PctMgrHmState using Census Divisions and Regions instead of states. The empirical estimates corroborate our earlier finding that mutual fund managers overweight their holdings in firms located close to their homes. In each of the regression specifications, the coefficient estimate on PctMgrHmDivision or PctMgrHmRegion is positively estimated and significant at the 5% level. Our finding that mutual fund managers overweight holdings close to their homes is not sensitive to alternative geographic classifications. 4.4 Alternative explanations Our result that fund managers overweight their home states and that this overweighting does not lead to superior performance is consistent with managers exhibiting a familiarity bias. An alternative explanation is that funds hire managers from certain areas of the country when they plan to build up their portfolio holdings from those areas. For example, a fund that invests primarily in technology stocks may be interested in hiring someone from California to be in charge of the selection of Silicon Valley firms. The analysis of home-state holdings around manager turnover casts doubt on this alternative explanation, since the selection story suggests that the fund is likely to begin building positions in the state prior to the arrival of the new manager. From Table 3, we see that this is not the case. We now provide additional evidence to rule out selection as the driving factor in home-state overweighting. We do so by testing whether managers who exhibit stronger home-state biases are selected from states where stocks performed well prior to those managers being hired. If funds want access to companies in certain states, then they are likely to choose those that have done well recently. If managers are selected to deliver investments in “hot” industries or states, then we would expect home-state overweighting to be higher after manager turnover events if home-state returns have been higher recently. To test this, we interact various state performance measures at the time of the manager’s hire with the MgrHmDum in our regression from column 1 of Table 5. The results of these tests are displayed in Table W3 in the Online Appendix. We use oneand three-year state portfolio returns in excess of the risk-free rate and CAPM alphas. To increase the power of the tests, we limit the sample to managers in their first four years of tenure with the fund, but using the entire sample

managers’home states and their holdings. We find that funds hold portfolios that are 18% closer to their managers’ home states than the holdings of the benchmark portfolio in the same Morningstar category (results are available upon request).

2593

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2593

2563–2601

The Review of Financial Studies / v 25 n 8 2012

leads to similar results. In all four specifications, the coefficient estimate on the interaction term is not significantly different from zero. Mutual funds are not selecting managers from “hot” states to gain access to firms in these states. Besides selection, another potential alternative explanation is the notion of “state pride.” While state pride would also imply that managers overweight their home states—especially when they have spent more time there—and that this overweighting is not driven by information, we showed earlier that our findings are robust to various distance measures and are not confined to state borders. Additionally, it is not clear that managers of funds with fewer resources would be more likely to overinvest in home-state securities if state pride were the motivating factor. 5. Conclusion Whether familiarity plays a role in investment decisions has long been of interest to economists. Although previous research has documented a familiarity bias in the investment choices of individual investors, ours is the first article to document such a bias among professionals. We show that the average fund in our sample overweights stocks from its managers’ home states. When new managers are hired, they quickly tilt their portfolios toward stocks from their home states. The bias is stronger among managers who are less experienced, spent more time in their home state, or work for funds with fewer resources. We also show that managers who overweight their home-state firms do not deliver superior performance on these holdings, confirming that the overweighting is not driven by information. Finally, we find that the cost borne by investors in funds managed by the most biased managers is underdiversification. We estimate that home-state familiarity accounts for about 87 bps of the portfolio of the average actively managed, domestic-equity mutual fund in our sample. While 87 bps may seem modest, U.S. mutual fund assets totaled $11.8 trillion in 2010, 30% of which were actively managed domestic stock funds (Investment Company Institute 2011). Our results imply that domestic equity mutual funds allocate roughly $31 billion worth of capital based on familiarity each year. Since we consider familiarity through only one channel—homestate stocks—and there is surely noise in our home-state measure, this likely is a conservative estimate. Our results have asset pricing implications. Cao, Han, Hirshleifer, and Zhang (2011) show that the premium investors require to trade unfamiliar stocks is increasing in the proportion of familiarity-biased investors in the market. Previous research has documented familiarity-based investing among individuals. We show that professional investors are also prone to this behavior. Since institutional investors account for the majority of trading in financial markets (Stulz 2007), finding evidence of a bias among them increases the likelihood that familiarity plays an important role in pricing assets.

2594

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2594

2563–2601

No Place Like Home

Our findings imply that some professional managers are subject to the same behavioral biases in their portfolio decisions as individual investors. To date, we know relatively little about how professional managers generate investment ideas. Our results suggest that familiarity is one heuristic that less-sophisticated managers use to generate such ideas. Appendix A. Data construction and variables This appendix provides additional detail of how we constructed our data, and variable definitions. Additional tables and figures are available in a separate Online Appendix. A.1 Identifying managers’ home states We identify the home state and age of each fund manager by locating them in the LexisNexis Public Records database. This database provides the state in which the Social Security Number (SSN) was issued, the year of issue, and the recipient’s month and year of birth. Prior to the Tax Reform Act of 1986, SSNs were typically assigned when a person got his or her first job or a driver’s license—usually at the age of 15 or so. We are therefore able to identify the place the person lived at this age, and we assume that this is where the manager grew up. Since SSNs are assigned to immigrants when they immigrate for work, we assume a person who obtained their SSN at an age of 22 or older is an immigrant and has no home state. This is the case for 109 (5.2%) of our identified managers. Our matching process proceeds as follows: 1. We begin with a search of the database for the full name of each fund manager in our sample. This step uniquely identifies 1,138 fund managers. If the search does not provide a unique result, we continue to step 2. 2. Morningstar sometimes includes the year in which the fund manager graduated from college. We use this to estimate the age of the manager by assuming the manager was between 18 and 24 years old when graduating. (We use this conservative window to avoid misidentifying managers.) This additional age information narrows the matches, allowing us to identify an additional 243 managers in this step. 3. For managers not uniquely identified in steps 1 or 2, we include the additional filter that the manager must have lived at some point in the state where the fund complex (his or her employer) is based. An additional 547 managers are identified in this step. 4. Finally, it is sometimes possible to determine the home state of a manager even if there is no unique match to a record in the LexisNexis database. This occurs when all possible matches have the same home state. In this case, we can identify which state the manager is from, but we cannot determine the manager’s age (because we don’t know which of the possible matches is correct). This step allows us to identify the home states of an additional 181 managers. Together, these steps allow us to uniquely identify 2,109 of the 4,236 managers in our original sample. None of the main results in the article is sensitive to restricting our sample to those matches identified only in Step 1. A manual comparison of a small sample of our matches with information gleaned from online biographical sources confirms that our identification strategy works well. A.2 Identifying fund adviser locations We identify the location of fund advisers from Form N-SAR, which is a required semiannual SEC filing. The form records the mutual fund family (“Registrant”) and all advisers and subadvisers for

2595

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2595

2563–2601

The Review of Financial Studies / v 25 n 8 2012

each fund (“Series”) within the fund family. A fund may have many advisers, and different funds within a family may have different advisers. The N-SAR captures all of this information, as well as the business address of each adviser. The N-SAR filing is identified by the SEC’s Central Index Key (CIK), which applies to all funds within a fund family. To identify which fund within a family is associated with which advisers during each quarter, we download all N-SAR filings for each fund family and extract the fund name and adviser zip code for each fund. We then use a name-matching algorithm to match funds reported in the N-SARs to our data. When using adviser data, we restrict our sample to those cases where there is no ambiguity in the name matching. This reduces our sample by about 34%. A.3 Variable definitions PctMgrHmStatei,s,t is the ratio of the number of managers of fund i from state s to the total number of managers of fund i during quarter t. MgrHmStateDumi,j,s,t is a dummy variable that is one if state s is the home state of manager j of fund i during quarter t. MorningstarBMWti,s,t is the average portfolio weight in geographic classification s (state, division, or region) of all funds within the same nine-box Morningstar category (US Large Blend, US Large Growth, US Large Value, US Mid-Cap Blend, US Mid-Cap Growth, US Mid-Cap Value, US Small Blend, US Small Growth, or US Small Value) as fund i during quarter t. MFHQStatei,s,t is a dummy variable that takes a value of one if the mutual fund management company of fund i is headquartered in state s during quarter t and is zero otherwise. Value is a dummy variable that is one if the mutual fund is categorized by Morningstar as a value fund. Growth is a dummy variable that is one if the mutual fund is categorized by Morningstar as a growth fund. SmallCap is a dummy variable that is one if the mutual fund is categorized by Morningstar as a small-cap fund. LargeCap is a dummy variable that is one if the mutual fund is categorized by Morningstar as a large-cap fund. FundSizeQuin is equal to the fund’s TNA quintile minus one. Funds are grouped into quintiles based on their total TNA each quarter. Assets under management are calculated from the Thomson’s holding data. FamSizeQuin is equal to the fund’s fund family TNA quintile minus one. Fund families are grouped into quintiles based on their total TNA each quarter using the universe of funds covered by the CRSP Mutual Fund Database. TeamManaged is a dummy variable that is equal to one if the fund is managed by more than one manager during quarter t and is zero otherwise. AgeGTMed is a dummy variable that is one if the age of the manager is greater than the median in the sample. Data on age is hand-collected from the LexisNexis online public records database. ExperienceGTMed is a dummy variable that is one if the mutual fund management experience of the manager is greater than the median in the sample. Experience is calculated as the number of years from the manager’s first day as a portfolio manager at any fund covered by the Morningstar database. HmTenureGTMed is a dummy variable that is one if the number of years that the manager lived in his home state is greater than the median in the sample. A particular manager’s home tenure is equal to his age if the manager’s home state matches the state in which the mutual fund is headquartered. If the two states do not match, then if the manager attended

2596

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2596

2563–2601

No Place Like Home

college in the same state as his home state, the age at which the manager graduated from his degree program is considered the manager’s home tenure. If the manager did not attend college in his home state and does not work for a fund headquartered in his state, then the manager is assumed to have left the state four years prior to obtaining a degree at an institution outside his home state. MgrCollState is a dummy variable that is one if the manager of the fund went to college in the state. Data on the names of managers’ colleges are from Morningstar, and we manually match college names to states. IvyLeague is a dummy variable that is one if the manager of the fund has a degree from an Ivy League institution. S&P 500 is a dummy variable that is one if the stock is included in the S&P 500 stock index. SalesGTMed is a dummy variable that is one if the sales of the firm are greater than the median in the sample. NumAnalystsGTMed is a dummy variable that is one if the number of analysts who cover the stock is greater than the median in the sample. AdvertisingExpGTMed is a dummy variable that is one if the advertising expense of the firm is greater than the median in the sample. StateBias is the difference between a fund’s actual portfolio weight in a state and the portfolio weight of the fund’s Morningstar benchmark in the state during the same time period. MFAdvStatei,s,t is a dummy variable that takes a value of one if the mutual fund adviser of fund i is headquartered in state s during quarter t and is zero otherwise. HmStateExRet1yr is the average monthly value-weighted portfolio return in excess of the risk-free rate of companies in the manager’s home state over the one year prior to the manager’s hire date. HmStateExRet3yr is the average monthly value-weighted portfolio return in excess of the risk-free rate of companies in the manager’s home state over the three years prior to the manager’s hire date. HmStateAlpha1yr is the CAPM alpha of the value-weighted portfolio return of companies in the manager’s home state estimated over the one year prior to the manager’s hire date. HmStateAlpha3yr is the CAPM alpha of the value-weighted portfolio return of companies in the manager’s home state estimated over the three years prior to the manager’s hire date. References Alexander, G. J., G. Cici, and S. Gibson. 2007. Does Motivation Matter When Assessing Trade Performance? An Analysis of Mutual Funds. Review of Financial Studies 20:125–50. Bhattacharya, U., and P. Groznik. 2008. Melting Pot or Salad Bowl: Some Evidence from U.S. Investments Abroad. Journal of Financial Markets 11:228–58. Cao, H. H., B. Han, D. Hirshleifer, and H. H. Zhang. 2011. Fear of the Unknown: Familiarity and Economic Decisions. Review of Finance 15:173–206. Carhart, M. M. 1997. On the Persistence of Mutual Fund Performance. Journal of Finance 52:57–82. Chevalier, J., and G. Ellison. 1999a. Are Some Mutual Fund Managers Better Than Others? Cross-sectional Patterns in Behavior and Performance. Journal of Finance 54:875–99. Chevalier, J., and G. Ellison. 1999b. Career Concerns of Mutual Fund Managers. Quarterly Journal of Economics 114:389–432. Cohen, L. 2009. Loyalty-based Portfolio Choice. Review of Financial Studies 22:1213–45.

2597

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2597

2563–2601

The Review of Financial Studies / v 25 n 8 2012

Cohen, L., A. Frazzini, and C. Malloy. 2008. The Small World of Investing: Board Connections and Mutual Fund Returns. Journal of Political Economy 116:951–79. Cohen, R. B., C. Polk, and B. Silli. 2010. Best Ideas. Working Paper, Harvard University. Coval, J. D., and T. J. Moskowitz. 1999. Home Bias at Home: Local Equity Preference in Domestic Portfolios. Journal of Finance 54:2045–73. Coval, J. D., and T. J. Moskowitz. 2001. The Geography of Investment: Informed Trading and Asset Prices. Journal of Political Economy 109:811–41. Cremers, M. K., and A. Petajisto. 2009. How Active Is Your Fund Manager? A New Measure That Predicts Performance. Review of Financial Studies 22:3329–65. Daniel, K., M. Grinblatt, S. Titman, and R. Wermers. 1997. Measuring Mutual Fund Performance with Characteristic-based Benchmarks. Journal of Finance 52:1035–58. French, K. R., and J. M. Poterba. 1991. Investor Diversification and International Equity Markets. American Economic Review 81:222–26. Goetzmann, W. N., and A. Kumar. 2008. Equity Portfolio Diversification. Review of Finance 12:433–63. Greenwood, R., and S. Nagel. 2009. Inexperienced Investors and Bubbles. Journal of Financial Economics 93:239–58. Grinblatt, M., and M. Keloharju. 2001. How Distance, Language, and Culture Influence Stockholdings and Trades. Journal of Finance 56:1053–73. Gruber, M. J. 1996. Another Puzzle: The Growth in Actively Managed Mutual Funds. Journal of Finance 51:783–810. Hong, H., and L. Kostovetsky. 2012. Red and Blue Investing: Values and Finance. Journal of Financial Economics 103:1–19. Hong, H., J. D. Kubik, and J. C. Stein. 2004. Social Interaction and Stock-market Participation. Journal of Finance 59:137–63. Huberman, G. 2001. Familiarity Breeds Investment. Review of Financial Studies 14:659–80. Hwang, B. 2011. Country-specific Sentiment and Security Prices. Journal of Financial Economics 100:382–401. Investment Company Institute. 2011. 2011 Investment Company Fact Book. http://www.icifactbook.org. Ivkovi´c, Z., and S. Weisbenner. 2005. Local Does as Local Is: Information Content of the Geography of Individual Investors’ Common Stock Investments. Journal of Finance 60:267–306. Jensen, M. C. 1968. The Performance of Mutual Funds in the Period 1945–1964. Journal of Finance 23:389–416. Jin, L., and A. Scherbina. 2011. Inheriting Losers. Review of Financial Studies 24:786–820. Kang, J.-K., and R. A. M. Stulz. 1997. Why Is There a Home Bias? An Analysis of Foreign Portfolio Equity Ownership in Japan. Journal of Financial Economics 46:3–28. Korniotis, G. M., and A. Kumar. Forthcoming. Tall versus Short: Height, Lifelong Experiences, and Portfolio Choice. Journal of Finance. Kumar, A. 2009. Who Gambles in the Stock Market? Journal of Finance 64:1889–1933. Lazarsfeld, P. F., and R. K. Merton. 1954. Friendship as a Social Process: A Substantive and Methodological Analysis. In Freedom and Control in Modern Society, Morroe Berger, Theodore Abel, and Charles H. Page, eds. New York: Van Nostrand, 18–66. Malmendier, U., and S. Nagel. 2011. Depression Babies: Do Macroeconomic Experiences Affect Risk-taking? Quarterly Journal of Economics 126:373–416. Morse, A., and S. Shive. 2011. Patriotism in Your Portfolio. Journal of Financial Markets 14:411–40.

2598

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2598

2563–2601

No Place Like Home

Parwada, J. T. 2008. The Genesis of Home Bias? The Location and Portfolio Choices of Investment Company Start-ups. Journal of Financial and Quantitative Analysis 43:245–66. Pirinsky, C., and Q. Wang. 2006. Does Corporate Headquarters Location Matter for Stock Returns? Journal of Finance 61:1991–2015. Pomorski, L. 2009. Acting on the Most Valuable Information: Best Idea Trades of Mutual Fund Managers. Working Paper, University of Toronto. Seasholes, M., and N. Zhu. 2010. Individual Investors and Local Bias. Journal of Finance 65:1987–2010. Seru, A., T. Shumway, and N. Stoffman. 2010. Learning by Trading. Review of Financial Studies 23:705–39. Shive, S. 2010. An Epidemic Model of Investor Behavior. Journal of Financial and Quantitative Analysis 45:169–98. Strong, N., and X. Xu. 2003. Understanding the Equity Home Bias: Evidence from Survey Data. Review of Economics and Statistics 85:307–12. Stulz, R. M. 2007. Hedge Funds: Past, Present, and Future. Journal of Economic Perspectives 21:175–94. Sulaeman, J. 2008. Do Local Investors Know More? Evidence from Mutual Fund Location and Investments. Working paper, Southern Methodist University. Tesar, L. L., and I. M. Werner. 1995. Home Bias and High Turnover. Journal of International Money and Finance 14:467–92. Wermers, R. R. 2005. Is Money Really “Smart”? New Evidence on the Relation Between Mutual Fund Flows, Manager Behavior, and Performance Persistence. Working paper, University of Maryland. Yonker, S. E. 2010. Geography and the Market for CEOs. Working Paper, Indiana University.

2599

[10:29 2/7/2012 OEP-hhs075online.tex]

Page: 2599

2563–2601