Do financial experts make better investment decisions?

Accepted Manuscript Do Financial Experts Make Better Investment Decisions? Andriy Bodnaruk, Andrei Simonov PII: DOI: Reference:

S1042-9573(14)00058-8 http://dx.doi.org/10.1016/j.jfi.2014.09.001 YJFIN 682

To appear in:

Journal of Financial Intermediation

Received Date:

4 July 2013

Please cite this article as: Bodnaruk, A., Simonov, A., Do Financial Experts Make Better Investment Decisions?, Journal of Financial Intermediation (2014), doi: http://dx.doi.org/10.1016/j.jfi.2014.09.001

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Do Financial Experts Make Better Investment Decisions?

Andriy Bodnaruk*

Andrei Simonov**

*

University of Notre Dame; ** Michigan State University, Gaidar Institute and CEPR. Please address all correspondence to Andrei Simonov, 321 Eppley Center, Michigan State University, East Lansing, MI 48824, phone: 517- 884-0455, e-mail: [email protected].

We appreciate helpful comments and suggestions from Shane Corwin, Steve Dimmock, Gur Huberman, Terry Odean, Paul Schultz, Murillo Campello (the editor), two anonymous referees, and seminar participants at University of Notre Dame, conference participants at the CFS-EIEF Conference on Household Finance and DePaul Behavioral finance conference “People & Money”. .

Do Financial Experts Make Better Investment Decisions?

Abstract

We provide direct evidence on the effect of financial expertise on investment outcomes by analyzing private portfolios of mutual fund managers. We find no evidence that financial experts make better investment decisions than peers: they do not outperform, do not diversify their risks better, and do not exhibit lower behavioral biases. Managers do much better in stocks for which they have an information advantage over other investors, i.e., stocks that are also held by their mutual funds. More experienced managers seem to be aware of the limitations to their investment skills as they increase their holdings of mutual fund-related stocks following poor performance of their portfolios. Our results suggest that there are limits to the value added by financial expertise.

JEL Classification: G02, G11, G23 Keywords: individual investors, investor sophistication, financial expertise, mutual fund managers

1

The general consensus in the academic literature is that a vast majority of individual investors would be better off putting their money in an index fund. 1 Poor investment decisions by individual investors are often blamed on their lack of financial sophistication, defined as the ability to avoid making investment mistakes (Calvet, Campbell, and Sodini, 2009). Indeed, sophistication has been related to higher stock market participation (Mankiw and Zeldes, 1991, Haliassos and Bertaut, 1995, Vissing-Joergensen, 2003, Christiansen, Schroter-Joense, and Rangvid, 2008, Grinblatt, Keloharju, and Linnainmaa, 2011); broader portfolio diversification (Goetzman and Kumar, 2008, Calvet et al., 2007); better performance (Seru, Shumway, and Stoffman, 2010, Grinblatt et al., 2012); and reduced behavioral biases (Feng and Seasholes, 2005, Calvet et al., 2009). Investors may avoid mistakes by having high cognitive abilities and /or by having financial expertise, where the latter is defined as a comprehensive knowledge of financial markets based on prolonged experience through practice and education. While one does not preclude the other, proxies, used in the literature to identify more financially sophisticated investors, appear to be good at capturing investors’ general intelligence, but do not seem to measure financial expertise well.2 In this paper we provide direct evidence on the effect of financial expertise on investment outcomes. We identify a group of individual investors who have the extensive knowledge of finance attained through prior training and day-to-day experience with financial markets: mutual

1

For an overview of empirical findings on individual investors’ behavior see, for example, Barber and Odean (2011). 2 Proxies of financial sophistication used in the literature include (disposable) income and wealth (Dhar and Zhu, 2006, Vissing-Joergensen, 2003, Calvet et. al, 2007, 2009), portfolio diversification (Goetzman and Kumar, 2008, Grinblatt and Keloharju, 2001), prior investment experience (Goetzman and Kumar, 2008, Nicolosi, Peng, and Zhu, 2009, Seru, Shumway, and Stoffman, 2010), educational attainment (Christiansen et al., 2008, Calvet et. al, 2007, 2009), investment in more complex financial instruments (Genesove and Mayer, 2001, Goetzman and Kumar, 2008), and IQ (Grinblatt et al. 2011, 2012).

2

fund managers.3 We compare private investment decisions by these financial experts to those made by individual investors which are similar to them along a number of socio-economic characteristics, but presumably lack financial expertise. We observe personal portfolios of 84 mutual fund (MF) managers in Sweden, as well as the portfolios of their mutual funds and peer individual investors. We have for these managers information on their real estate, total wealth, and personal characteristics. We find that financial experts do not exhibit superior security-picking ability in their own portfolios. Private investments of fund managers perform on par with investments of investors similar to them in terms of age, sex, education level, income, and wealth. Even more striking, mutual funds managers’ investments perform more poorly than the private investments of the wealthiest 1% of investors. Our financial experts (mutual fund managers) are likely to have superior access to information and/or analysis about certain companies gained in the course of their work. To disentangle the effect of financial expertise from that of information advantage we conduct the following robustness analysis. We reason that if a security is held by the mutual fund run by manager, ceteris paribus he is more likely to have information advantage in that security. To control for these information differences, we split portfolios of managers into positions that are held by the mutual fund of the manager (MF-related), and those which are not (non-MF-related) and investigate them separately. 4 We find that non-MF-related investments of managers significantly

3

We must emphasize that in this paper we consider mutual fund managers as individual investors with a high level of financial expertise and use them as a study group to investigate the relation between financial expertise and the quality of investment decisions of individual investors. Whether fund managers have skills to create value for mutual fund investors is intentionally left outside of our analysis as well as any cross-sectional performance variation within the manager group. 4 It is subject to a debate whether financial expertise is defined only by the knowledge of financial markets or it also involves access to information that is not available to general public. By analyzing overall private portfolios of fund managers and sub-portfolios of their non-MF-related positions we provide evidence on the quality of experts’ decisions which is consistent with both interpretations.

3

underperform their MF-related investments. This suggests that a part of overall managerial performance should be credited to access to mutual fund’s resources. Financial experts also do not appear to be better than peers at diversifying risks. Managers invest a higher percentage of their wealth in mutual funds, but hold a similar number of individual stocks and exhibit similar levels of portfolio concentration. As a result, the Sharpe ratios of their investments are similar to those of control investors. Nor do we find evidence that financial experts are less affected by behavioral biases. There is some evidence that managers exhibit a lower (in fact, negative) disposition effect, but only in MF-related positions. They also turn over their portfolios as often as control group. Interestingly, financial experts seem to be aware of their limitations with regard to their investment abilities. Underperforming managers, especially the more experienced of them, subsequently increase their allocations to MF-related positions. A one-standard deviation lower past-year portfolio return is related to an 8.4% (or 37.7% relative to unconditional mean) higher probability of holding a position in MF-related stocks and a 9.1% (or 124.6% relative to the sample mean) larger share of MF-related positions in the total value of the portfolio in a subsequent year. Still, only about 22% of manager-year observations in our sample hold any position that is also held by their mutual fund. Our results can be best summarized in the following way: financial expertise is of little value for investors in the top decile by investable wealth. It is plausible that marginal effect of financial expertise on investment decisions is trivial for these investors, but is of larger importance for less well-off individuals.5 Our results, nevertheless, provide important insights as they demonstrate that there are limits to the value added by financial expertise.

5

Recent evidence suggests that less sophisticated investors may not be able to use expert knowledge to their benefit. Bhattacharya, Hackethal, Kaesler, Loos, and Meyer (2012) find that retail investors, who receive free unbiased

4

We make several contributions. First, we contribute to the literature on individual investor decisions. Investor sophistication has been linked to better investment outcomes. The role of financial knowledge so far has been underexplored. Our results demonstrate that for highly sophisticated investors (defined as the top decile of population by financial wealth), expertise in finance does not improve investment decisions. Second, our results help to explain the stylized fact that many high net worth individuals do not seek the services of investment advisors, but instead prefer to invest on their own. 6 Wealthy investors appear to be as good individual investors as professional asset managers. Low value added by investment advice for wealthy investors does not seem to justify the (hefty) fees. There are several potential criticisms concerning the validity of our results related to the small size of financial experts’ sample, size of their equity portfolios, the potential for social interactions between experts and control investors, restrictions on trading by mutual fund managers in their private accounts etc. To streamline our exposition we address these issues comprehensively in an online Appendix. The rest of the paper is organized as follows. In section 1 we describe our data sources and present descriptive statistics of characteristics of managers and control groups of investors. Section 2 presents results. Section 3 concludes.

investment advice from finance professional and follow it, significantly improve efficiency of their portfolios. However, they also report that a) only about 5% of investors accept the offer of (free) advice; and b) investors who need the advice most (e.g., those with poor past performance) are least likely to seek and follow it. 6 A recent survey of 1,000 young (under-50) wealthy (with investable assets of over half a million dollars) investors conducted by Cisco finds 30% of these investors do not have a financial advisor. Similarly, a Spectrem Group study of investors under age 45 finds that over 25% of wealthy investors with 100K-1MM and 12% of millionaires do not use any advisors. Between 40% and 67% of advisor-less high-net worth investors in these surveys believe they can do a better job of investing than a professional advisor.

5

1. Data and Descriptive Statistics 1.1 Data Sources Our data come from several sources. For each mutual manager as well as for control groups of investors (described below) we have information on age, wealth, location, and portfolio holdings on the level of individual domestic and foreign securities (stocks, bonds, mutual funds). Similarly, for each mutual fund we have data on portfolio holdings and fund characteristics. These combined data sources are available for 6 years, from July 2001 till June of 2007. Below we describe these data sources in more detail.

Individual investor portfolio holdings Data on individual investor portfolio composition are obtained from Swedish government records (KURU). These data are available because Sweden levies a wealth tax. To collect this tax, the government assembles records of financial assets, including mutual funds that are held outside defined-contribution pension accounts. The records go down to the individual security level and are based on statements from financial institutions that are verified by taxpayers. The dataset also provides information on real estate holdings, bank accounts, positions in bonds and derivatives, and loans for the entire population. For a more detailed description of this dataset see Calvet et al. (2007).

Individual investor characteristics LINDA (Longitudinal INdividual DAta for Sweden) is a register-based longitudinal dataset that is a joint endeavor between the Department of Economics at Uppsala University, the National Social Insurance Board (RFV), Statistics Sweden, and the Ministries of Finance and Labor. LINDA consists of a large representative panel of households over the period 1966 — 2000. For each year, information on all family members of the sampled individuals is added to 6

the dataset. The sampling procedure ensures that the data are representative for each year. Moreover, the same family is traced over time. This provides a time-series dimension that, in general, is lacking in surveys based on different cohorts polled over time. The variables include individual characteristics (sex, age, marital status, country of birth, citizenship, year of immigration, place of residence at the parish level, education, profession, employment status); housing information (type and size of housing, owner and rental status, single-family or multi-family dwelling, year of construction, housing value); and tax and wealth information. In particular, the income and wealth tax registers include information on labor income, capital gains and losses, pension contributions, taxes paid, and taxable wealth. A detailed description of the dataset is provided by Edin and Fredriksson (2000).7

Mutual fund managers We obtain the names of all mutual fund managers in Sweden as of March 30, 2006 from mutual fund websites. These as well as various other open sources were used to identify manager’s age (often as precise as the exact date of birth). We then manually match mutual fund managers to their tax records, which we obtain from the Swedish Tax Authority. Those records include information on income, wealth, debt, and household structure. Individual-level information on tax returns is publicly available in the Swedish Tax Authority offices. Some individuals can request their records to be withheld, citing, for example, security concerns. There are 5 managers in our sample for whom tax records were concealed. A more significant reason for the loss of observations is that some fund managers have generic names. As a result, there were several people with identical names and years of birth. When we were not

7

These data are available on the website http://linda.nek.uu.se/

7

able to match unique fund managers to their tax records, we dropped these managers from our sample. We are able to successfully match 128 of 218 managers (59%) to their tax records. Next, we link fund managers to their individual portfolio holdings and characteristics in KURU database (described above). Unfortunately, not all managers have information reported in this database. We also drop managers who managed their fund for less than 36 months. We end up with 84 usable portfolios. One potential criticism of our identification of fund managers and their portfolios is that our approach is overly restrictive; indeed, our final sample includes only managers for whom we have full confidence as to the validity of all the relevant data. It could be argued that by making “educated guesses” we would be able to increase the sample size and improve the statistical power of our tests. This approach, however, would likely result in misidentification of some managers (“false positives”), which would tend to bias our results in the direction of finding no differences in the quality of decisions by financial experts and non-experts; i.e., the results that we find. An alternative way to increase the size of our sample of experts would be to adopt a broader definition of “financial experts.” That is, we could say everyone who works in a financial industry is a financial expert. Indeed, since our data specify individuals’ occupations at the industry level, this would yield thousands of “experts.” The obvious drawback of this approach is that the vast majority of people working in a financial industry, such as banks’ front-office personnel, marketing professionals, or tax accountants, do not have a significant edge in knowledge of and experience with financial markets. We, thus, focus on mutual fund managers, as these are the individual investors, who are most likely to have financial expertise.

8

While we are able to observe only a subsample of managers, we believe that it is not biased toward including less skilled managers. First, more successful managers tend to be in charge of larger mutual funds. The 84 managers in our sample represent 38.5% of the population of Swedish managers; they manage 30% of the positions in Swedish-registered mutual funds; and one out of every three individual investors invests in mutual funds managed by investors in our sample. Second, the largest Swedish financial newspaper, Dagens Industri, together with Morningstar recognizes the top three mutual funds in a variety in categories on a yearly basis. While the number of categories varies from year to year, about 44% of these prizes went to managers in our sample. Yet another criticism of our identification of managers is related to a fact that our sample suffers from survivorship bias; i.e., poorly performing managers leave the asset management industry early and, as a result, are excluded from our sample. This, however, should bias our results toward finding that managers are better investors than control investors. We do not find evidence of that. Moreover, if survivorship bias is a problem, our results indicate that managers should perform more poorly than their peers. In that sense, our “no-result” is an upper limit estimate of managers’ ability.

Mutual fund holdings By Swedish law, the positions of mutual funds registered in Sweden are reported quarterly to the Swedish Financial Supervisory Authority, or Finansinspektionen. Since 2001 these filings have been publicly available via Internet. The database includes fund and fund family identification, name, country and currency codes, and ISIN of security, number of shares held, and market value of such holding as well as exchange rate applicable (if any).

9

Firm-level information and other data Individual security returns (including dividends) are derived from several databases including COMPUSTAT Global, Datastream, Moneymate, and SIX Trust (Swedish analog of CRSP).

Performance Benchmarks The question of asset pricing benchmark is complicated by the fact that Sweden is a small open economy that represents a small part of the world market portfolio. Using data on portfolio holdings from the Coordinated Portfolio Investment Survey held by the IMF, Sercu and Vanpee (2008) show that home bias in Sweden is among the smallest among countries with sizable stock markets; 59.4% of equity is held in domestic equity in December of 2005. To put this number into perspective, local bias was 82.2% for the United States, 91.9% for Japan, and 68.8% for France.8 Thus, the relevant benchmark should include not only domestic market, but also have an international component. We acknowledge it in our analysis by using two kinds of benchmarks. The first one is purely domestic, and is built using Morgan Stanley Capital Indices. We use MSCI Sweden index as a proxy for the market return. The difference between returns on MSCI Sweden Small and Micro Cap Index and MSCI Sweden Large Cap is used as a proxy for SMB factor. Similarly, we use the difference between returns on MSCI Sweden Value and MSCI Sweden Growth as a proxy for HML factor. We decided to stick with three-factor domestic model as Rouwenhorst (1998) showed that Sweden is one of very few markets in which the momentum factor is insignificant. For the second set of benchmarks, we use a new set of international global benchmarks developed by Fama and French (2012). The global factors include 23 countries in 4 regions:

8

We do not argue with the fact that for the US the extent of home bias is actually relatively smaller (as the US represents 40% of the world market capitalization). However, the use of national indices for the US or Japan is more justified than the use of national benchmarks for Sweden.

10

Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Hong Kong, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Singapore, Spain, Sweden, Switzerland, United Kingdom, and the United States. Using a methodology similar to the original Fama-French (1993) and Carhart (1997) papers, they construct four factors that explain the returns in the local markets well. These data are available from Kenneth French’s website. We use returns denominated in Swedish Kronas (SEK).9 Three-month STIBOR rate is used as a proxy for the risk-free rate. For exposition purposes, we report results only for the world-factors adjusted performance analysis. The results for domestic factors adjusted returns are similar. 1.2. Comparison Groups for Mutual Fund Managers To isolate the effect of financial expertise on investment we identify a control group of individual investors who are similar to mutual fund managers in terms of socio-economic characteristics, but do not possess similar financial expertise. Matching is done in 1999, i.e., before any analysis of investment decisions is executed; we preserve the control group throughout the sample period. We match by wealth, labor and capital income (closest individual within 10% bounds), age, sex, family status, and educational achievement. We consider two versions of the match. In the first one we do not impose constraints on peers’ industry of occupation. In the second one we limit the matched sample to people who are not employed in the financial industry (11% of matches in the first match). The biggest industries represented in the matched sample are business services (21% in the first match, 24% in the second), trade (20%, 23%), and mining and manufacturing (17%, 18%). Overall, control investors appear to be similar to fund managers along a large number of characteristics, but differ 9

To transform Fama-French (2012) returns denominated in US dollars into SEK we used data on exchange rate from Swedish Central Bank.

11

from them in terms of day-to-day exposure to financial markets. Generally, the results for both matches are very similar and for the sake of brevity we report the results only for the first match. We also compare investment decisions of fund managers to decisions of the wealthy investors in Sweden. We consider two groups: top 1% of population by financial wealth (defined as the sum of the value of common stock and mutual fund holdings), and the 95th to 99th percentiles in terms of financial wealth.10 The use of the 95th to 99th financial wealth percentiles bracket as a control group is of particular importance. Investors with the highest wealth levels – i.e., the top 1% – are most likely to use personalized financial advisory services. Thus, it could be argued that portfolio decisions for this group of individual investors are still made by financial experts, albeit externally hired. Access to private financial services is, however, very exclusive. For example, one of the largest Swedish banks, SEB, requires customers to have at least 500,000 Euros in investable wealth to obtain access to its private financial services. This is similar to, for example, Vanguard’s requirements ($500,000). Investors in the 95th to 99th financial wealth percentiles group predominantly do not meet this criterion: mean (median) value of equity portfolios for investors in this bracket is about 1,244,000 (801,000) Swedish Krona (or roughly 134,000 (90,000) USD), which suggests that most investors in this bracket are very unlikely to use personalized investment advice. To consider this from a different perspective, financial advisors typically charge fees of about 1% of assets under management (although, as mentioned above, these usually apply to much larger portfolios) – a 1% fee on the portfolio of this size would be about $1,340 ($900). Quality

10

We also used groups of investors in the 90th to 95th financial wealth percentiles and 75th to 90th percentiles, but the results for this group are qualitatively the same as for 95th-99th percentile group and are omitted for brevity.

12

financial advice is likely to cost considerably more.11 Hence, comparing the investment decisions of managers to investors in the 95th to 99th financial wealth percentiles allows us to rule out the possibility that our findings could be explained by investors with similar levels of financial expertise making decisions in both cases. 1.3. Characteristics of Managers and Matched Peers We present descriptive statistics of managerial income variables in Panel A of Table 1. Average yearly labor income of Swedish mutual fund managers is almost 1 million SEK and is growing at about 10.9% per year. This compares favorably with the average salary of nonmanual workers in the private sector (approximately 371,000 SEK) and its growth (3.7%) as reported by Statistical Central Bureau (SCB) of Sweden. Capital income, with a mean (median) of 397,000 (107,000) SEK, is another significant source of managerial income. Average (median) assessed wealth is about 3.3 (2.2) times yearly salary. Real estate represents about 2/3 of managerial wealth. These means and medians indicate that managers (and matched investors) on average are in the top 0.5% of the population in terms of labor and capital income. Managers also keep a considerable amount of wealth in individual stocks (Panel B). The average (median) value of their stock portfolios is 766,000 (136,000) SEK. Managers also hold a greater number of stocks – average (median) of 6.5 (4.0) – than the average investor.12 Managers share an average of three positions with the mutual funds they work (median of two), which represents approximately 1/3 of their individual equity portfolio. In Table 2 we report descriptive statistics for the bank, debt, and real estate holdings of mutual fund managers and the matched group. Relative to peers, the median fund manager keeps 11

It is unclear whether financial advisors aim to create value for their clients in the first place. Chalmers and Reuter (2014) find that investors using the services of financial advisors underperform those who do not. Mullainathan, Noeth, and Schoar (2012) document that advisers fail to mitigate behavioral biases of their clients. Both studies suggest that advisors may have agendas conflicting with clients’ objectives. 12 Bodnaruk (2009) reports the average number of stocks per Swedish individual investor is about 2.1.

13

a similar amount of wealth in bank deposits, but obtains higher interest rates: median equalweighted (value-weighted) interest rate is 28 bp (26 bp) higher. Managers also have less debt. About 5% of managers do not report debt at all, compared to less than 1% of matched individuals. The median manager has about 521,971 SEK in two credit lines, whereas the median matched individual has 668,991 SEK in three credit lines. At the same time, managers pay significantly lower interest rates. For median equal-weighted (valueweighted) interest rates, the difference is 69 bp (83 bp). This difference cannot be explained by the risk associated with higher debt, as the difference of roughly 150,000 SEK is trivial compared to the average income of about 1 million SEK13. Managers hold very few bonds. Only 6% of manager-year observations have any bonds associated with them, and monetary values are low. The median value of bond holdings, conditional on holding bonds, is SEK 24,252 (vs. median equity holdings of SEK 135,771). For coupon bonds, it is interesting to note that coupon rates are significantly higher for managers than for matched individuals (median difference between 5.00% and 2.57%). While we are not able to identify credit ratings for these bonds, there are no defaulted bonds in our sample. Sweden also has a market for lottery bonds. These bonds, for which coupon payment is determined by lottery, have a high level of idiosyncratic risk (Green and Rydqvist, 1997). The fraction of lottery bonds among managers is significantly lower than among matched individuals (12% vs. 30%). It seems that managers do not like additional idiosyncratic risk associated with these bonds. Overall, it appears that managers are more knowledgeable than peers about the array of available financial opportunities when it comes to instruments with certain outcomes. They are 13

Managers employed by banks (like SEB or Handelsbanken) might be able to obtain lower interest rates via employee discounts. We investigate whether there is any difference between managers associated with banks and those who are not and find no evidence of such discounts.

14

able to deposit money at higher interest and borrow at lower interest rates. They also avoid idiosyncratic risk. We now proceed to investigate whether managers have better ability to predict future stock returns.

2. Results 2.1. Performance 2.1.1. Returns by Position We start by calculating monthly returns by position for financial experts – mutual fund managers – and a less financially savvy group of matched investors as well as two groups of wealthy investors. Lacking return data on debt and real estate assets, we focus on positions in individual equity and mutual funds. From summary statistics provided in Table 3, Panel A, we can see managers hold 36,203 monthly positions in stocks and mutual funds (1). About 9.8% (3,534 observations) of these positions are shared with the mutual fund of the manager (MF-related, 2), and 32,669 positions are not shared, non-MF-related (3). Additionally, we break down MF-related positions into those, that are not held by any other mutual fund in the same fund family, MF-NMFF (986 positions, 2a) and those, that are, MF-MFF (2,548 positions, 2b). MF-NMFF positions are in securities on which managers demonstrate such strong conviction that they include them in their private portfolios and in portfolios of their mutual funds even though other managers in the same fund family do not share their enthusiasm; they represent managers’ “best ideas” in the spirit of Cohen, Polk, and Silli (2010). We also identify positions in managers’ portfolios that are not shared with their own mutual funds, but which are held by other funds in the group, NMF-MFF (886 positions, 3a).

15

Matched investors hold somewhat larger number of monthly positions than fund managers overall – 42,777 (4). Naturally, investors in wealthiest groups hold considerably more positions (over 68 million for the top one percentile group (5) and close to 150 million for 95th to 99th percentiles group (6)). Since we have only annual snapshots of investor portfolios, we make an assumption that a position recorded on June 30, was held from that day until the end of June of the next year.14 We calculate monthly return on each position for every investor, and equally weight them within each investor group. The difference in average monthly returns per position for managers and matched peers [(1) (4)] is economically small – about 9bp per month – and statistically insignificant. When we decompose managers’ portfolios into MF-related and non-MF related positions, however, we observe that managers significantly outperform their peers in MF-related positions [(2) - (4)] by 48bp per month. This outperformance is particularly pronounced when we consider managers’ “best ideas,” i.e., positions that they share with their own mutual fund, but not with other funds in the family (MF-NMFF). In contrast, managers do not deliver higher returns per position than peers when their positions are not shared with mutual funds [(3) - (4)] – the difference in performance is not only statistically insignificant, but also economically trivial at 5bp per month. Overall, managers underperform the wealthiest group of investors [(1) - (5)] – by 14bp per month, and do as well as investors in 95th to 99th wealth percentile [(1) - (6)]. It is worth pointing out, however, that on average managers are among the top 1% of the wealthiest people in Sweden. Also, it is interesting to note that positions that managers share with other funds in the family, but not with their own fund, NMF-MFF, (3a) exhibit the poorest performance of all groups considered. 14

A similar assumption is made by Calvet et al. (2007, 2009).

16

2.1.2. Investor Portfolio Returns We explore the performance of fund managers further in Table 3, Panel B. Here we report descriptive statistics of monthly portfolio returns by investor. As before, we consider positions only in equities and mutual funds. We value-weight positions within each investor’s portfolio and then average these returns across investors. Consistent with prior results, managers outperform neither matched peers nor either group of wealthy individuals. Moreover, the median test shows that top 1% outperformed managers by economically and statistically significant 24 bp per month. 2.1.3. Portfolio Returns by Investor Group Panel C reports average portfolio returns by investor group. We aggregate all positions of managers, peers, and groups of wealthy investors in value-weighted portfolios. We also consider portfolios of MF-related positions (with a further division into MF-NMFF and MF-MFF positions) and non-MF-related positions (and a subset of NMF-MFF positions) by managers. As before, we find no evidence of outperformance of managers over matched investors overall [(1) - (4)], but there is weak evidence that managers outperform less financially knowledgeable investors in the stocks they share with their mutual funds [(2) - (4)]. Similarly, we find no evidence that managers outperform wealthy investors. Consistent with prior results, there is some evidence that the top 1% of investors actually outperform financial experts. Consistent with analysis of returns by position, the portfolio based on NMF-MFF positions performs the worst. 2.1.4. Portfolio Returns by Investor Group: Adjusting for Risk

17

To account for risk, we perform multifactor portfolio analysis. In Table 4, Panel A, we report one- and four-factor-adjusted returns for portfolios of managers and matched investors using the Fama and French (2012) four world factors as benchmarks. Neither managers nor their peers are able to beat the market. While in both cases alphas are positive, they are not statistically significant at conventional levels. More important, managers do not outperform peers. In fact, going long in the managers’ portfolio and shorting the peers portfolio yields negative alpha; although economically very small (minus 6 bp per month) and statistically insignificant. It is plausible that managers have no superior skills with respect to individual securities, but possess better market timing abilities. We explore this possibility by estimating Treynor-Mazuy (1966) and Henriksson-Merton (1981) market timing regressions. For the Treynor-Mazuy regressions, we add the quadratic term of the excess return on the market. In the HenrikssonMerton regressions, the additional term is the maximum value of zero and excess return on the market. The results do not support a conjecture that fund managers excel in market timing.15 In Panel B we compare the abnormal performance of managers relative to wealthy investors. The results are even less flattering for managers; going long in portfolios of managers and shorting portfolios of the wealthiest 1% of investors delivers negative alpha of between 28 bp and 45 bp per month which is statistically significant at 5% level. Performance attribution analysis shows that the outperformance of the wealthiest 1% is actually related to stock selection and not market timing. The abnormal performance of managers relative to investors in the 95th to 99th wealth percentiles is also negative, although neither statistically nor economically significant.

15

We acknowledge that lack of evidence on market timing ability of managers could be attributed to us missing transaction level data (Puckett and Yan, 2011).

18

So far our results indicate that, in general, managers do not possess superior stock picking ability. It could of course be the case that they do have this skill with respect to a subset of securities, e.g., securities about which they have superior information stemming from their workplace. We therefore investigate performance of portfolios of MF-related and non-MFrelated positions (Table 5, Panel A). Once again, we find that managers do much better investing in securities that they share with their mutual funds than in securities they do not; the difference in monthly abnormal returns is 54 bp per month for the one-factor model and 58 bp per month for multifactor models, both are significant at the 10% level. Interestingly, the outperformance of the MF-related portfolio appears to be driven by market timing. The economic effect of market timing is 1.1% per month for Treynor-Mazuy estimation and 0.9% per month for the Henriksson-Merton estimation. In Panel B we dissect MF-related positions of mutual fund managers into finer categories and investigate the performance of resulting portfolios relative to the portfolio of managers’ positions that are not shared with their own mutual funds, but that are held by other funds in the group, NMF-MFF. We split positions that are shared with MF of the manager into those that are not held by any other mutual fund in the same fund family, MF-NMFF, and those that are, MF-MFF. Managers perform much better – by 91 bp per month – in the positions that they share with their mutual funds than in positions that are not held by their funds, but that are held by other funds in their mutual fund family. This outperformance is limited, however, to managers’ “best ideas,” i.e., positions that are shared with their own funds, but not other funds in the family (MFNMFF).

19

Overall, the results provide little evidence that fund managers have better stock picking skills than control groups of investors, particularly when they cannot capitalize on resources of their employers.16 2.2. Diversification and Sharpe Ratio Financial expertise should not only help investors achieve better performance, but also enable them to more properly diversify their risks and, thus, to gain more for the amount of risk taken. To explore diversification of managers’ portfolios, we consider three measures: number of positions, portfolio concentration, and share of individual stocks in the total value of the portfolio of stocks and mutual funds. We calculate these measures for each investor separately and then equal-weight them at a group level. The results, presented in Table 6, indicate that managers allocate a lower fraction of their portfolios to individual equities than matched investors and the wealthiest 1% of investors, but a similar fraction as investors in the 95% to 99% wealth bracket. However, they hold a similar number of stock positions, and their portfolios exhibit a similar degree of concentration as peers. Moreover, according to these two measures they are underdiversified compared to both groups of wealthy individuals. As a result, portfolio Sharpe ratios for managers are not only indistinguishable from Sharpe ratios for matched investors, but also both statistically and economically much lower than for wealthy investors. These results suggest that managers are no better at diversifying their risks and obtaining superior reward per unit of risk than non-financial experts.

16

We also use information about star mutual fund managers to compare investment outcomes across fund managers. We split managers into two groups: those ever recognized as a top 3 manager in their category and those which were not. We do not find any evidence that star fund managers make better investment decisions in their own portfolios than less prominent fund managers.

20

2.3. Behavioral Biases Might managers be less likely to fall victim to behavioral biases? We examine two behavioral biases: the disposition effect and excessive portfolio turnover (both in stocks and mutual funds, and in stock positions only). In our analysis of the disposition effect we follow Calvet et al. (2007, 2009) who use the same data sources. As in Odean (1998) and Dhar and Zhu (2006), we use proportions of stock gains (losses) realized during the year. A gain (loss) in a particular stock is counted as realized if the investor sells some (but not necessarily all) of its holdings of the stock. The investor’s proportion of gains (losses) realized, PGRi,t (PLRi,t), is then defined as the number of winning (losing) stocks with realized gains divided by the total number of winning (losing) stocks. The disposition effect is then measured by the difference: Disp = PGRi,t – PLRi,t. The properties of our data (absence of purchase prices) make it necessary to follow Calvet et al. (2009) and base our classification on return on the index. We classify a stock as a winner if it has a higher return than the Fama-French World Index during the year and as a loser if it underperforms the index. The results are reported in Table 7. We find evidence that fund managers exhibit a lower disposition effect – both overall and in stocks only – than matched investors, but not than groups of wealthy investors. When we split portfolios of managers into MF-related and non-MF-related positions, we observe that the lower disposition effect comes from MF-related holdings, while there is no statistically discernible difference in the disposition effect in non-MF-related positions of managers and overall portfolios of comparison groups.

21

The results on portfolio turnover are similar. For overall portfolios and in non-MF-related positions, we find that managers engage in similar or even higher levels of portfolio turnover than comparison groups. As before, the MF-related positions are exceptions; managers reshuffle these to a significantly lower degree. To summarize, managers do not seem to be less prone to behavioral biases than non-financial experts with similar characteristics. 2.4. Portfolios of Managers and Portfolios of Mutual Funds of Managers So far, our results indicate that managers make better decisions with respect to positions they share with their mutual funds. To explore this further, we first analyze the choice of fund-related positions, and then investigate whether managers are aware of the limitations on their investment abilities. We start with a basic frequency distribution of managerial holdings conditional on whether manager’s mutual fund also holds a security. In Table 8, Panel A, we observe that managers are about 3.6 times more likely to pick a stock that is held by their mutual fund than a stock that is not. The difference is economically and statistically significant for both mean and median tests and robust to different corrections for clustering. This finding is supported in multivariate (probit and Tobit) analysis (Panel B). Here we use the sample of managers and matched investors to ask about the likelihood and the size of the position that they take in a given security. We perform this analysis on the set of all stocks and mutual funds that are owned by managers, matched investors, or mutual funds at any point in time (these include both Swedish domestic securities and foreign securities). The dependent variable is a dummy that takes a value of 1 if an investor holds a position in the security at the end of June of the subsequent year and 0 otherwise (probit) or a logarithm of

22

10 of a value of a position (Tobit) at the corresponding date. Control variables include Match (a dummy indicating that an investor is a matched investor rather than a manager); MF Holds (a dummy indicating that the mutual fund of a manager holds a position in the security); MF-NMFF Holds (a dummy indicating that the mutual fund of a manager holds a position in the security, but none of the funds in the same mutual fund family does), MF-MFF Holds (a dummy indicating that both the mutual fund of the manager and at least one other fund in the same mutual fund group holds a security); Domestic (a dummy if a security is issued by a domestic company); MF (a dummy indicating a security is a mutual fund); PopProb (fraction of the overall population of Swedish individual investors holding the security); and Age and Age Squared (measuring the non-linear effect of the age of the investor). MF Holds, MF-MFF Holds, and MFF-NMFF Holds are equal to zero for matched investors. If a position is held by the mutual fund of a manager, it is 0.3% (or 65% relative to sample mean) more likely to appear in the manager’s own portfolio; shared positions are, on average, 2.8 times larger than non-MF-related positions. At the same time, managers are about equally likely to invest in the stocks that are held only by their mutual funds as in the stocks that are held not only by their own funds but also by other funds in the same group; the amounts of money they invest in such positions are also comparable. This result is puzzling, given that MF-NMFF positions perform (at least economically) better than MF-MFF positions (see Table 3 and Table 5, Panel B). Managers also prefer to invest in domestic securities, mutual funds, and securities widely held by general population, but avoid volatile assets. They also load more strongly on stocks and mutual funds that performed well recently and those that continue to do well in the near future.

23

Next, we explore how managers reshuffle their portfolios in response to the evidence about their investment ability. If a manager performs poorly this would indicate that his stock picking skills are limited. At the same time, MF-related positions deliver better performance than nonMF-related positions. Henceforth, underperforming managers should rebalance their portfolios toward MF-related stocks. We relate a fraction of the portfolio invested in MF-related positions to the past year return on their portfolio in Table 9. Among control variables we include a dummy Positiont-1 indicating whether a manager has any position in MF-related stocks in the previous period; lag and lead return on manager’s portfolio; HiAge dummy that takes a value of one if a manager is older than the median manager in our sample (38 years old); and interactions between these variables. We also use time and manager fixed effects. In Panel A we find that some managers seem to be aware of the limitations to their investment skills. Managers who perform relatively poorly subsequently increase their allocations to MFrelated positions. A one standard deviation lower past year portfolio return results in an 8.4% higher probability of subsequently holding a position (or 37.7% relative to unconditional mean) in MF-related stocks and a 9.1% increase in the share of MF-related positions in the total value of the portfolio (or 124.6% relative to the sample mean). From Panel B it also appears that these are the more experienced managers – with abovemedian age – who both hold a larger fraction of their portfolios in MF-related positions and rebalance their portfolios in response to past performance. In contrast, the relation between past portfolio return and allocation in MF-related stocks is at best weak for less experienced managers.

24

Even though the effect seems to be very significant economically, still only about 22% of manager-year observations in our sample hold any MF-related position at any point in time. This seems like a very low number which raises the concern that at least some fund managers may be restricted from holding fund-related securities. We, therefore, investigated the legal and regulatory guidelines on fund managers’ ability to hold fund-related securities and discussed this issue with several fund managers directly. While certain restrictions are in place, none of them appears to be a significant deterrent against managers holding a large fraction of their portfolios in MF-related stocks.17

3. Conclusions We provide direct evidence as to the effect of financial expertise on investment. We identify a group of individual investors who have the extensive knowledge of finance attained through prior training and day-to-day experience with financial markets; mutual fund managers – and compare their investment decisions to decisions made by investors with similar characteristics, but without comparable financial expertise. We find no evidence that our financial experts make better investment decisions: they do not outperform, do not diversify their risks better, and do not exhibit lower behavioral biases. Managers do much better in the stocks they share with their mutual funds, but less than a quarter of them have any of these mutual fund-related positions. Overall, our results demonstrate that for high net worth investors expertise in finance does not improve investment decisions.

17

Insider trading laws in Sweden prohibit front-running, so the fund managers are not allowed to take positions in stocks with any knowledge that their mutual funds are about to take positions in these stocks as well. Many mutual funds also do not allow managers to make short-term (less than one month) round-trip transactions in securities owned by the fund. Neither of these restrictions bans managers from investing in a security already owned by the mutual fund long-term.

25

References: Barber, Brad, and Terrance Odean, 2011, The Behavior of Individual Investors, working paper. Bhattacharya, Utpal, Andreas Hackethal, Simon Kaesler, Benjamin Loos, and Steffen Meyer, 2012, Is Unbiased Financial Advice to Retail Investors Sufficient? Answers from a Large Field Study, Review of Financial Studies 25, 975-1032. Bauer, Rob, Mathijs Cosemans, and Piet Eichholtz, 2008, Option Trading and Individual Investor Performance, Journal of Banking and Finance 33, 731-746. Bodnaruk, Andriy, 2009, Proximity Always Matters: Local Bias When the Set of Local Companies Changes, Review of Finance 14, 629-656. Brown, Jeffrey, Zoran Ivkovic, Paul Smith, and Scott Weisbenner, 2008, Neighbors Matter: Causal Community Effects and Stock Market Participation, Journal of Finance 65, 1509-1531. Calvet, Laurent, John Campbell, and Paolo Sodini, 2007, Down or Out: Assessing the Welfare Costs of Household Investment Mistakes, Journal of Political Economy 115, 707-747. Calvet, Laurent, John Campbell, and Paolo Sodini, 2009, Measuring the Financial Sophistication of Households, American Economic Review 99, 393-398. Carhart, Mark M., 1997, On Persistence in Mutual Fund Performance, Journal of Finance 52, 57–82. Chalmers, John, and Johnathan Reuter, 2014, What is the Impact of Financial Advice on Retirement Portfolio Choice and Outcomes?, working paper Cohen, Randy, Christopher Polk, and Bernhard Silli, 2010, Best Ideas, Working paper. Christiansen, Charlotte, Juanna Schroter-Joense, and Jesper Rangvid, 2008, Are Economists More Likely to Hold Stocks? Review of Finance 12, 465-496. Dhar, Ravi, and Ning Zhu, 2006, Up Close and Personal: An Individual Level Analysis of the Disposition Effect, Management Science 52, 726-740. Edin, Per-Anders, and Peter Fredriksson, 2000, LINDA – Longitudinal Individual Data for Sweden, working paper. Ehrenberg, Ronald G., and Bognanno, Michael L, 1990a, Do Tournaments Have Incentive Effects? Journal of Political Economy 98(6), 1307-1324. Ehrenberg, Ronald G., and Bognanno, Michael L, 1990b, The incentive effects of tournaments revisited: Evidence from the European PGA tour, Industrial and Labor Relations Review 43(3), 74-88. Fama, Eugene F., and Kenneth R. French, 1993, Common risk factors in the returns on stocks and bonds, Journal of Financial Economics 33, 3-56. Fama, Eugene F., and Kenneth R. French, 2012, Size, Value, and Momentum in International Stock Returns, Journal of Financial Economics, forthcoming. Feng, Lei, and Mark Seasholes, 2005, Do Investor Sophistication and Trading Experience Eliminate Behavioral Biases in Financial Markets? Review of Finance 9, 305-351.

26

Genesove, David, and Christopher Mayer, 2001, Loss Aversion and Seller Behavior: Evidence from the Housing Market, Quarterly Journal of Economics 116, 1233-1260. Goetzmann, William, and Alok Kumar, 2008, Equity Portfolio Diversification, Review of Finance 12, 433-463. Green, Rick, and Kristian Rydqvist, 1997, The Valuation of Non-Systematic Risks and the Pricing of Swedish Lottery Bonds, Review of Financial Studies 10, 447-479. Grinblatt, Mark, and Matti Keloharju, 2001, What Makes Investors Trade? Journal of Finance 56, 589-616. Grinblatt, Mark, Matti Keloharju, and Juhani Linnainmaa, 2011, IQ and Stock Market Participation, Journal of Finance 66, 2121-2164. Grinblatt, Mark, Matti Keloharju, and Juhani Linnainmaa, 2012, IQ, Trading Behavior, and Performance, Journal of Financial Economics 104, 339-362. Haliassos, Carol, and Michael Bertaut, 1995, Why Do So Few Hold Stocks? The Economic Journal 105, 1110-1129. Heaton, John, and Lucas, Deborah, 2000, Portfolio Choice and Asset Prices: The Importance of Entrepreneurial Risk, Journal of Finance 55(3), 1163-98. Henriksson, Roy, and Robert Merton, 1981, On the Market Timing and Investment Performance of Managed Portfolios II - Statistical Procedures for Evaluating Forecasting Skills, Journal of Business 54, 513-533. Hong, Harrison, Jeffrey Kubik, and Jeremy Stein, 2004, Social Interaction and Stock Market Participation, Journal of Finance 59, 137-163. Lazear, Edward P., and Sherwin Rosen, 1981, Rank-Order Tournaments as Optimum Labor Contracts, Journal of Political Economy 89(5), 841-864. Mankiw, Gregory and Stephen Zeldes, 1991, The Consumption of Stockholders and Nonstockholders, Journal of Financial Economics 29, 97-112. Matthews, Peter H., Paul M. Sommers, and Francisco J. Peschiera, 2007, Incentives and Superstars on the LPGA Tour, Applied Economics 39(1), 87-94. Melton, Michael and Thomas S. Zorn, 2000, An Empirical Test of Tournament Theory: The Senior PGA Tour, Managerial Finance 26(7) 16-32. Mullainathan, Sendhil, Markus Noeth, and Antoinette Schoar, 2012, The Market for Financial Advice: an Audit Study, working paper Nicolosi, Gina, Liang Peng, and Ning Zhu, 2009, Do Individual Investors Learn from Their Trading Experience? Journal of Financial Markets 12, 317-336. Odean, Terrance, 1998, Are Investors Reluctant to Realize Their Losses? The Journal of Finance 53, 1775–1798. Orszag, Jonathan M., 1994, A New Look at Incentive Effects and Golf Tournaments, Economics Letters 46(1), 77-88. Puckett, Andy and Xuemin Yan, 2011, The Interim Trading Skills of Institutional Investors, Journal of Finance 66, 601-633. 27

Rhoads, Thomas A., 2008, Are Prizes in Professional Golf Too High?, Towson University working paper. Rouwenhorst, K. Geert (1998), International Momentum Strategies, The Journal of Finance, vol. 53, no. 1, pp. 267-284. Sercu, Piet and Rosanne Vanpee, 2008, Estimating the Costs of International Equity Investments, Review of Finance 12, 587-634. Seru, Amit, Tyler Shumway, and Noah Stoffman, 2010, Learning by Trading, Review of Financial Studies 23, 705-739. Shmanske, Stephen, 1992, Human Capital Formation in Professional Sports: Evidence from the PGA Tour, Atlantic Economic Journal 20(3), 66-80. Taubman, Paul, and Terence Wales, 1974, Higher Education and Earnings: College as an Investment and Screening Device, McGraw-Hill, New York Treynor, Jack and Kay Mazuy, 1966, Can Mutual Funds Outguess the Market? Harvard Business Review 44, 131-136. Vissing-Joergensen, Anette, 2003, Perspectives on Behavioral Finance: Does “Irrationality” Disappear with Wealth? Evidence from Expectations and Actions, NBER Macroeconomic Manual.

28

Table 1: Descriptive Statistics of Managers Panel A reports managerial income variables based on LINDA data over 1998-2002. Statistics for labor and capital income, wealth, and growth rate of labor income are based on Carroll and Samwick (1998) decomposition. Panel B reports stock portfolio-related variables. We report value of portfolio and number of positions for each manageryear. All monetary values are in Swedish Kroners (SEK). Panel A: Managerial Income Variables N

Mean

Median

Std Dev

IQR

Labor Income

551

992,292

865,182

643,757

707,411

Capital Income

147

397,326

107,412

887,432

385,783

Wealth (based on wealth tax)

157

3,327,263

1,794,738

5,003,563

1,744,523

Income Growth Rate

438

0.109

0.134

0.503

0.320

Panel B: Stock Portfolios of Managers N

Mean

Median

Std Dev

IQR

Value of Portfolio

504

766,196

135,771

3,045,111

407,943

Number of Positions

504

6.486

4.000

6.266

7.000

Value of MF-related Portfolio

103

159,767

75,037

257,791

173,029

Number of MF-related Positions

103

2.952

2.000

2.396

3.000

Share of MF-related Stocks

103

0.346

0.335

0.238

0.360

29

Table 2: Bank, Debt, and Real Estate Holdings: Managers vs. Matched Sample In Panel A we report variables related to bank holdings of mutual funds’ managers vs. matched sample. We report fractions of managers and peers who report bank holdings, total deposits (in SEK), average number of bank accounts, equal- and value-weighted deposit rates (calculated for each bank account as total interest payments over the year divided by end-of-year bank account balance). Data are from KURU register and are available for 1999, 2000, 2003-2006. In Panel B we report variables related to debt. We report total debt amount (in SEK), number of lines, equal- and value-weighted debt interest measures (calculated for each bank account as total interest payments over the year divided by end-of-year debt account balance). In Panel C we report holdings of real estate based on KURU 2003-6. In Panel D we report the descriptive statistics for bond holdings. We report number of positions per year-person conditional on holding bonds, size of individual positions (in SEK), coupon in percent (conditional on being reported), fraction of lottery and short-term bonds, and lowest denomination for bonds held. Panel A: Bank Holdings % Reported Bank Holdings Total Deposits # of Bank Accounts Deposit Rate (equal-weighted) Deposit Rate (value-weighted) Panel B: Bank Loans % Reported Debt Total Debt # of Open Debt Lines Loan Rate (equal-weighted) Loan Rate (value-weighted) Panel C: Real Estate % Reported Real Estate Total Value # of Items Reported Panel D: Bond Holdings % with Bond Positions # of Positions Bond Holdings Coupon Lottery Bonds Fraction of Short-term (<18 months) Lowest Denomination

N 407 407 407 407 N 526 526 526 526 N 122 122 N 56 74 31 74 74 74

Managers Mean 93.14% 181,527 1.585 0.0208 0.0204 Mean 95.29% 873,684 2.688 0.0566 0.0368 Mean 58.37% 573,311 1.049 Mean 6.14% 1.321 129,064 5.429 0.122 0.392 2,543

Median

N

62,361 1.000 0.0126 0.0119

497 497 497 497

Median

N

521,971 2.000 0.0328 0.0327

658 658 658 658

Median

N

362,098 1.000

58 58

Median

N

1.000 24,252 5.000 0.000 0.000 1,000

136 267 74 267 267 267

30

Peers Mean 97.84% 326,674 1.497 0.0188 0.0181 Mean 99.25% 990,087 3.027 0.0647 0.0437 Mean 36.25% 542,406 1.086 Mean 15.31% 1.963 79,033 4.414 0.296 0.4157 4,470

t-test Median 76,294 1.000 0.0098 0.0093 Median 668,991 3.000 0.0397 0.0410 Median 333,177 1.000 Median 1.000 44,500 2.570 0.000 0.000 1,000

t-stat -3.42 -2.96 1.47 1.28 1.33

p-value 0.001 0.003 0.142 0.201 0.183

Wilcoxon Z-stat p-value -3.53 0.000 -1.52 0.128 0.87 0.384 2.24 0.025 1.95 0.051

t-stat -4.11 -1.23 -3.62 -1.73 -3.07

p-value 0.000 0.220 0.000 0.083 0.000

Z-stat -4.35 -3.43 -4.20 -3.80 -6.66

p-value 0.000 0.001 0.000 0.000 0.000

t-stat 4.20 0.30 -0.76

p-value 0.000 0.767 0.450

Z-stat 4.10 0.48 -0.56

p-value 0.000 0.629 0.576

t-stat -6.37 -2.940 1.80 1.74 -3.68 -0.37 -3.66

p-value 0.000 0.004 0.076 0.087 0.000 0.713 0.000

Z-stat -6.30 -0.813 -1.3873 2.21 -3.03 -0.37 -3.53

p-value 0.000 0.417 0.165 0.027 0.003 0.713 0.000

Table 3: Performance of Managers, Matched Peers, and Wealthy Investors: Raw Returns In Panel A we report univariate statistics per position of average monthly return for managers (line 1) vs. matched sample (line 4) and other comparison groups (line (5) -Top 1% of equity holders, line (6)- 95-99% of equity holders). We also report separately positions of managers in securities that are owned by the funds they managed (MF positions, line 2), and all other positions (non-MF positions, line 3). In the former case, we also look separately at the positions that are held by managers and no other fund in the same family (MF-NMFF positions, line 2a), and by managers in our sample and other managers in their family (MF-MFF positions, line 2b). In the latter case, we report separately the results for the positions that are held by managers and other mutual funds that belong to the same family, but not by the fund of the manager (NMF-MFF positions, line 3a). We also report the results of mean and median tests between groups. Panel B reports average monthly portfolio returns by investor. Panel C provides statistics on average monthly portfolio return by investor group. Panel A: Returns by Position N

Mean

Median

(1) Managers’ Positions

36,203

0.0090

0.0085

(2) MF Positions

3,534

0.0129

986

0.0176

2,548

0.0111

(2a) MF-NMFF Positions (2b) MF-MFF Positions (3) Managers: non-MF Positions (3a)NMF-MFF Positions (4) Matched Individuals

t-test

p-value

Z-stat

p-value

(1)-(4)

1.38

0.169

2.12

0.034

0.0130

(2)-(4)

2.67

0.008

2.24

0.025

0.0162

(2)-(3a)

6.98

0.000

6.50

0.000

0.0113

(2b)-(3a)

6.76

0.000

5.71

0.000

32,669

0.0086

0.0081

(3)-(4)

0.75

0.454

1.71

0.087

886

-0.0184

-0.0091

(2a)-(3a)

6.92

0.000

6.46

0.000

42,777

0.0081

0.0058

(2)-(3)

-2.37

0.018

-1.70

0.089

(5) Top 1%

68,373,031

0.0104

0.0104

(1)-(5)

-2.82

0.005

-4.47

0.000

(6) 95-99%

149,869,103

0.0090

0.0073

(1)-(6)

0.05

0.956

-0.14

0.900

Panel B: Investor Portfolio Returns Difference with Managers N

Mean

Median

t-stat

p-value

Z-stat

p-value

Managers’ Positions

84

0.0086

0.0081

Matched Individuals

94

0.0077

0.0078

0.45

0.656

0.08

0.934

Top 1%

85,746

95-99%

375,215

0.0092

0.0105

-0.31

0.757

-2.29

0.0216

0.0071

0.0076

0.84

0.403

0.32

0.749

31

Panel C: Portfolio Returns by Investor Group N

Mean

Median

t-test

p-value

Z-stat

p-value

(1) Managers’ Positions

72

0.0059

0.0112

(1)-(4)

-0.52

0.605

0.06

0.951

(2) MF Positions

72

0.0108

0.0115

(2)-(4)

1.67

0.099

0.13

0.897

72

0.0114

0.0084

(2)-(3a)

1.92

0.059

1.66

0.097

(2a) MF-NMFF Positions (2b) MF-MFF Positions (3) Managers: non-MF Positions (3a)NMF-MFF Positions

72

0.0069

0.0043

(2b)-(3a)

1.08

0.286

0.56

0.572

72

0.0053

0.0061

(3)-(4)

-1.00

0.321

-0.35

0.729

72

0.0014

-0.0097

(2a)-(3a)

1,74

0.087

1.66

0.097

(4) Matched Individuals

72

0.0064

0.0095

(2)-(3)

1.82

0.073

1.15

0.249

(5) Top 1%

72

0.0104

0.0104

(1)-(5)

-2.25

0.028

0.34

0.736

(6) 95-99%

72

0.0090

0.0073

(1)-(6)

-0.43

0.669

0.03

0.976

32

Table 4: Performance of Managers and Comparison Samples: Abnormal Portfolio Returns We report the results of CAPM and Fama-French-Carhart regressions of portfolio returns for portfolios of mutual funds managers, comparison samples, and differences between them. In Panels A, we report the results for managers, matched sample, and the difference between them. For the long-short portfolios we also report Treynor-Mazuy and Henriksson-Merton market timing regressions. In Panel B we report the results for comparison samples (Top 1% and 95-99% group by the value of the portfolio). Positions are value-weighted within investor portfolios and then equal-weighted across investors. There are 72 observation months. Panel A: Managers vs. Peers Managers Estimate t-stat

Peers estimate t-stat

estimate

t-stat

Difference Managers vs. Peers estimate t-stat estimate t-stat

estimate

t-stat

Intercept

0.0049

(1.40)

0.0053

(1.43)

-0.0005

(-0.49)

-0.0007

(-0.74)

-0.0007

(-0.49)

0.0006

(0.73)

FFWMKTRF

1.1021

(12.72)

1.1304

(12.11)

-0.0283

(-1.13)

0.0183

(0.56)

0.0205

(0.51)

0.0294

(1.25)

FFWSMB

-0.0807

(-1.56)

-0.0805

(-1.55)

-0.0491

(-1.46)

FFWHML

-0.0476

(-0.72)

-0.0474

(-0.71)

0.0010

(0.02)

FFWWML

0.0874

(2.16)

0.0874

(2.15)

0.0501

(1.93)

-0.0063

(-0.09) 0.2824

(1.03)

Max(FFWMKTRF,0) FFWMKTRF2 Adj R2

0.6935

0.6722

0.0038

0.0417

33

0.0273

0.0370

Panel B: Managers vs. 95th and 99th Wealth Percentiles and Top 1% by Wealth

Intercept

Difference Managers vs. 95-99% estimate t-stat estimate t-stat

estimate

t-stat

Difference Managers vs. Top 1% estimate t-stat estimate t-stat

estimate

t-stat

-0.0004

(-0.40)

-0.0023

(-1.70)

-0.0018

(-1.47)

-0.0028

(-2.36)

-0.0044

(-2.80)

-0.0042

(-2.85)

FFWMKTRF

0.1665

(4.96)

0.1164

(2.88)

0.2002

(5.38)

0.0778

(1.98)

0.0349

(0.73)

0.1098

(2.50)

FFWSMB

0.0278

(0.52)

0.0231

(0.44)

0.0090

(0.17)

-0.0035

(-0.06)

-0.0075

(-0.12)

-0.0213

(-0.34)

FFWHML

-0.0707

(-1.03)

-0.0750

(-1.12)

-0.0663

(-0.99)

0.0044

(0.05)

0.0007

(0.01)

0.0085

(0.11)

FFWWML

-0.0929

(-2.23)

-0.0930

(-2.29)

-0.0941

(-2.30)

-0.0817

(-1.68)

-0.0818

(-1.70)

-0.0828

(-1.72)

0.1454

(2.11)

0.1244

(1.53)

0.8402

(1.94)

0.7975

(1.56)

Max(FFWMKTRF,0) FFWMKTRF2 Adj R2

0.499

0.524

0.519

0.164

34

0.180

0.182

Table 5: Performance of Mutual Fund-Related Stocks in Managers’ Portfolios We report the results of CAPM and Fama-French-Carhart regression of portfolio returns for different cuts of portfolios of mutual funds managers. In Panel A, we report the results for portfolio of securities that is held by both managers’ personal and mutual fund portfolio, portfolio of securities that is held by managers’ personal portfolio and not in mutual fund portfolio, and the difference between those. For the long-short portfolios we also report Treynor-Mazuy and Henriksson-Merton market timing regressions. In Panel B, we report the alpha of the strategy of going short in the securities that were held by managers and other mutual funds that belong to the same family, but not by the fund of the manager (NMF-MFF positions), and going long in (a) MF-related positions, or (b) positions that are held by managers’ private and mutual fund portfolio and another fund of the same fund management company(MF-MFF positions), or (c) positions that are held by managers’ private and mutual fund portfolio and are not held by another fund of the same fund management company(MF-NMFF positions). We report alphas for one and four-factor Fama-French international model. T-statistics is reported in parentheses. There are 72 observation months. Panel A: MF-related vs. non-MF-related Portfolios of Managers MF-related estimate t-stat

non-MF-related estimate t-stat

Difference: MF-related vs. non-MF-related t-stat estimate t-stat estimate t-stat estimate

estimate

t-stat

Intercept

0.0097

(1.81)

0.0043

(1.29)

0.0054

(1.80)

0.0058

(1.96)

-0.0038

(-1.13)

-0.0011

(-0.34)

FFWMKTRF

1.1977

(8.96)

1.0849

(13.09)

0.1128

(1.49)

0.0925

(0.95)

-0.0955

(-0.93)

0.2598

(2.56)

FFWSMB

-0.0042

(-0.03)

-0.0974

(-0.74)

-0.0974

(-0.67)

FFWHML

-0.0526

(-0.26)

0.0016

(0.01)

-0.0309

(-0.17)

FFWWML

-0.1581

(-1.31)

-0.1688

(-1.63)

-0.1638

(-1.47)

0.5763

(3.29) 4.1738

(3.53)

Max(FFWMKTRF,0) FFWMKTRF2 Adj R2

0.5276

0.7058

0.0169

0.0701

0.2159

0.2056

Panel B: MF-related Positions and Holdings by Other Funds in the Same Fund Family One-Factor Model estimate

t-stat

Four-Factor Model estimate

t-stat

Long

Short

MF Positions

NMF-MFF Positions

0.0091

(1.94)

0.0095

(2.03)

MF-MFF Positions

NMF-MFF Positions

0.0052

(1.05)

0.0057

(1.15)

MF-NMFF Positions

NMF-MFF Positions

0.0095

(1.78)

0.0097

(1.82)

35

Table 6: Diversification and Sharpe Ratio We report the results for measures of portfolio diversification (number of positions, Herfindahl index of portfolio concentration and share of individual stocks relative to overall value of portfolio of stocks and mutual funds) and Sharpe ratio for managers and matched investors (peers). We also report tests of differences between different groups. We used data for 2002:07-2007:06. Difference with managers t-test Wilcoxon t-stat p-value Z-stat p-value

N

Mean

Median

Share of stocks Managers Peers Top 1% 95-99%

84 96 81,825 348,452

0.450 0.639 0.686 0.446

0.385 0.747 0.828 0.383

-3.30 -6.35 0.10

0.001 0.000 0.923

-3.28 -5.37 0.36

0.001 0.000 0.717

Herfindahl Index Managers Peers Top 1% 95-99%

84 96 81,825 348,452

0.546 0.546 0.314 0.379

0.506 0.485 0.205 0.304

0.00 7.99 5.20

0.999 0.000 0.000

0.07 7.92 5.53

0.944 0.000 0.000

Number of positions Managers Peers Top 1% 95-99%

84 96 81,825 348,452

6.034 6.151 17.190 9.963

4.132 3.549 14.708 8.000

0.13 -18.25 -6.44

0.896 0.000 0.000

0.05 -10.49 -6.22

0.957 0.000 0.000

Sharpe Ratio Managers Peers Top 1% 95-99%

84 96 81,825 348,452

0.185 0.185 0.283 0.307

0.143 0.150 0.224 0.221

0.02 -2.86 -3.05

0.987 0.004 0.002

0.14 -4.43 -4.07

0.892 0.000 0.000

36

Table 7: Behavioral Biases We report the results for disposition effects (defined as PLG-PLR) and position turnover both for overall portfolio of stocks and mutual funds and for stocks only. In Panel A we report the data for overall’ managers’ portfolio and comparison samples. In Panel B we report the results for MF-related positions vs. nonMF-related positions for mutual funds managers. We also report tests of differences between different groups. The data used are for 2003-2006. Panel A: Managers vs. Peers

N

Mean

Median

Difference with managers t-test Wilcoxon t-stat p-value Z-stat p-value

Disposition Effect Managers Matched Peers Top 1% 95-99%

84 96 70,189 286,428

-0.0244 0.0381 -0.0137 -0.0028

-0.0116 0.0231 0.0000 0.0000

-1.69 -0.34 -1.34

0.093 0.733 0.181

-2.11 -0.48 -1.62

0.035 0.632 0.105

Disposition Effect: Stocks Only Managers Matched Peers Top 1% 95-99%

84 96 70,189 286,428

-0.0077 0.0515 -0.0034 0.0199

0.0048 0.0336 0.0000 0.0000

-1.92 -0.15 -1.45

0.056 0.878 0.148

-2.13 -0.55 -0.53

0.033 0.586 0.594

Turnover Managers Matched Peers Top 1% 95-99%

84 96 70,189 286,428

0.6169 0.5981 0.4981 0.5151

0.6129 0.6476 0.4848 0.5022

0.45 3.99 3.42

0.653 0.000 0.001

-0.24 4.59 3.83

0.815 0.000 0.000

Turnover: Stocks Only Managers Matched Peers Top 1% 95-99%

84 96 70,189 286,428

0.4300 0.4550 0.3878 0.2982

0.4314 0.4355 0.3215 0.2328

-0.50 1.49 3.44

0.621 0.137 0.001

0.68 1.66 3.35

0.499 0.098 0.001

37

Panel B: Managers: MF Related vs. non-MF Related Positions

Disposition Effect Disposition Effect: Stocks Only Turnover Turnover: Stocks Only

N 22 22 22 22

MF-related Mean -0.1720 -0.1720 0.3574 0.3574

Median -0.1422 -0.1422 0.3167 0.3167

N 84 84 84 84

38

Non-MF-related Mean -0.0115 0.0055 0.6353 0.4677

t-test Median -0.0116 0.0085 0.6500 0.5000

t-stat -2.37 -3.00 -4.08 -1.40

p-value 0.020 0.003 0.000 0.169

Wilcoxon Z-stat p-value -2.53 0.012 -2.97 0.003 -3.72 0.002 -2.34 0.020

Table 8: Portfolio Choice of Managers In Panel A we report the frequency distribution of managerial holdings conditional on their mutual fund holding the security. As universe of stocks we used all equity instruments that are owned by managers, their peers, or mutual funds. Overall we have 1991 equity instruments. In square brackets we report the percentage of overall sample. We also report the mean and median test of the difference between groups of stocks that are held by mutual funds vs. the rest. For mean and median tests we report the results without clustering, clustered over managers and clustered over firms. In Panel B we report the results of probit and Tobit regressions. The dependent variable is either a dummy which takes the value of 1 if a manager invests in a security and zero otherwise or a log10 of the value of the position. Match is a dummy equal to one if individual is in matched sample, and zero if individual is mutual fund manager. MF Holds a dummy equal to 1 if mutual fund of a manager holds the stock, and zero otherwise. MFNMFF Holds is a dummy equal to 1 if only mutual fund of a manager holds the stock, but not other funds in the same fund family, and zero otherwise. MF-MFF Holds is a dummy equal to 1 if mutual fund of a manager and other mutual funds in the same family holds the stock, and zero otherwise. MF Holds, MF-NMFF Holds, and MF-MFF Holds are zero for matched sample. Domestic is a dummy equal to 1 if the security is Swedish stock or mutual fund, and zero otherwise. MF is a dummy equal to one if security is a mutual fund, and zero otherwise. Age and Age2 are age of individual and square of age. PopProb is probability of holding of a particular security by general population. Lag(return) and Lead(return) are lagged and lead returns of a security over one year. StdDev(return) is standard deviation of stock return over 1 previous year. All probit coefficients are multiplied by 100. Standard errors are clustered at the firm level. Panel A: Frequency Distribution Mutual Funds Holds No Manager Does Not Hold

Yes

564,842

13,934

[ 99.57% ]

[ 98.45% ]

2,453

219

[ 0.43% ]

[ 1.55% ]

Statistics

p-value

Mean test (t-statistics, no clustering)

19.24

0.000

Mean test (t-statistics, clustering over managers)

2.51

0.014

Mean test (t-statistics, clustering over firms)

2.84

0.005

Wilcoxon median test (Z-statistics, no clustering)

19.23

0.000

Somers’ D test (Z-statistics, clustering over managers)

2.48

0.013

Somers’ D test (Z-statistics, clustering over firms)

3.14

0.002

Manager Holds Test of the difference:

39

Panel B: Multivariate Analysis Probit (1)

Tobit

(2)

(3)

(4)

(5)

(6)

dF/dX

t-stat

dF/dX

t-stat

dF/dX

t-stat

estimate

t-stat

estimate

t-stat

estimate

t-stat

Match

-0.284

(-1.48)

-0.279

(-1.49)

-0.279

(-1.49)

0.185

(0.88)

0.185

(0.89)

0.185

(0.89)

MF Holds

0.264

(4.09)

0.233

(3.91)

2.671

(2.97)

2.413

(2.77)

MF- NMFF Holds

0.251

(3.03)

2.142

(1.99)

MF-MFF Holds

0.231

(3.57)

2.458

(2.57)

0.268

(10.58)

5.713

(9.97)

5.784

(10.23)

5.784

(10.23)

Domestic

0.270

(10.34)

0.268

(10.58)

MF

0.107

(5.66)

0.070

(3.21)

0..070

(3.21)

1.921

(4.93)

1.190

(2.44)

1.190

(2.44)

Age

-0.014

(-0.88)

-0.011

(-0.75)

-0.011

(-0.75)

1.006

(4.90)

1.005

(4.89)

1.005

(4.89)

Age2 /100

0.014

(0.59)

0.011

(0.50)

0.011

(0.50)

-0.994

(-4.54)

-0.993

(-4.53)

-0.993

(-4.53)

PopProb

6.058

(9.66)

5.881

(9.68)

5.881

(9.67)

136.177

(9.61)

134.965

(9.82)

134.965

(9.82)

Lag(Return)

0.003

(5.49)

0.004

(5.78)

0.004

(5.78)

0.070

(5.39)

0.090

(5.78)

0.090

(5.78)

Lead(Return)

0.023

(3.72)

0.553

(3.90)

StdDev(Return) Year FE Individual FE 2

Pseudo R N

0.024

(3.72)

0.023

(3.72)

-0.271

(-2.97)

-0.271

(-2.97)

0.592

(3.93)

0.592

(3.93)

-7.720

(-3.31)

-7.720

(-3.31)

Y

Y

Y

Y

Y

Y

Y

Y

Y

Y

Y

Y

0.1978

0.1987

0.1987

0.1321

0.1332

0.1332

1256212

1256212

1256212

1256212

1256212

1256212

40

Table 9: Past Performance of Managers and Relative Share of MF-related Positions We report the determinants of taking positions in mutual funds- related stocks. In Panel A we report the results of baseline analysis. In Panel B we condition on manager age. We report findings of probit estimation (marginal effects and t-statistics) as well as the results of Tobit estimation for share of portfolio and logarithm of monetary value of mutual funds related holdings. We used lagged explanatory variable, Lag(return portfolio) and Lead(return(portfolio) – lagged (lead) one year returns of individual managers’ portfolio, HiAge (LoAge)– dummy that is equal to one if age of manager is higher or equal (lower) than median (38 years), and zero otherwise. Fixed effects are as reported in the tables. We used robust estimate adjusted for clustering over year.

41

Panel A: Baseline Analysis

Probability of Having a Position in MF-Related Stocks dF/dX

Positiont-1 Lag(Return Portfolio) Lead(Return Portfolio) Positiont-1 x Lag(Return Portfolio)

t-stat

dF/dX

t-stat

Share of Portfolio in MF-Related stocks Estimate

Pseudo R Nobs

Estimate

t-stat

Estimate

t-stat

Estimate

t-stat

0.070

(0.31)

0.020

(0.10)

0.101

(0.86)

0.118

(1.16)

-0.188

(-0.75)

-0.199

(-0.82)

-0.452

(-2.26)

-1.494

(-2.10)

-0.485

(-3.68)

-0.529

(-5.13)

-3.848

(-4.31)

-6.032

(-4.82)

1.471

(4.48)

1.078

(1.99)

0.416

(2.65)

0.340

(1.32)

4.411

(2.72)

3.860

(1.99)

-0.438

(-1.38)

-0.710

(-5.51)

0.384

(2.87)

0.347

(2.46)

2.909

(2.58)

2.187

(1.78)

0.384

(2.87)

0.347

(2.46)

2.117

(8.90)

2.083

(8.50)

σ 2

t-stat

Dollar Value in MF-Related Stocks

0.817

0.831

0.989

0.991

0.478

0.483

360

360

360

360

360

360

Year FE

N

Y

N

Y

N

Y

Individual FE

Y

Y

Y

Y

Y

Y

Panel B: Conditioning on Manager Age

Probability of Having a Position in MF-Related Stocks

Share of Portfolio in MF-Related stocks

Dollar Value in MF-Related Stocks

dF/dX

t-stat

dF/dX

t-stat

Estimate

t-stat

Estimate

t-stat

Estimate

t-stat

Estimate

t-stat

HiAge

0.103

(2.45)

0.114

(2.68)

0.101

(0.86)

0.118

(1.16)

-0.188

(-0.75)

-0.199

(-0.82)

HiAge x Lag(Return Portfolio)

-0.352

(-4.09)

-0.537

(-4.00)

-0.485

(-3.68)

-0.529

(-5.13)

-3.848

(-4.31)

-6.032

(-4.82)

HiAge x Lead(Return Portfolio)

0.288

(1.93)

0.267

(1.65)

LoAge x Lag(Return Portfolio)

-0.143

(-1.38)

-0.259

(-1.94)

0.416

(2.65)

0.340

(1.32)

4.411

(2.72)

3.860

(1.99)

LoAge x Lead(Return Portfolio)

0.347

(0.98)

0.336

(1.14)

0.384

(2.87)

0.347

(2.46)

2.909

(2.58)

2.187

(1.78)

0.384

(2.87)

0.347

(2.46)

2.117

(8.90)

2.083

(8.50)

χ2

p-val

χ2

p-val

χ2

p-val

χ2

p-val

5.72

0.004

4.47

0.012

1.28

0.279

1.39

0.251

σ Test of High Age interactions = Low Age interactions

Pseudo R2 Nobs

χ2 1.44

p-val 0.488

χ2 1.78

p-val 0.410

0.551

0.559

0.527

0.560

0.283

0.286

360

360

360

360

360

360

Year FE

N

Y

N

Y

N

Y

Individual FE

N

N

N

N

N

N

42