University of Southern California Dissertations and Theses

Mutual fund screening versus weighting

(USC Thesis Other)

Mutual fund screening versus weighting

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content *University of Southern California, Marshall School of Business, Department of Finance and Business
Economics (Roman.Skripnik.2019@marshall.usc.edu). 3670 Trousdale Pkwy, Los Angeles, CA 90089. I am
grateful to my committee members Wayne Ferson, Juhani Linnainmaa, Chris Jones, Fernando Zapatero
and Richard Sloan, to Richard Evans for valuable comments and help.

Mutual Fund Screening Versus Weighting

Roman Skripnik

Doctor of Philosophy (BUSINESS ADMINISTRTION)
FACULTY OF THE USC GRADUATE SCHOOL

August, 2019

2

Table of Contents

I. Introduction ............................................................................................................................ 4
II. The data .................................................................................................................................. 8
III. Selection of the benchmark index ........................................................................................ 11
IV. Methodology ......................................................................................................................... 12
A. Screening........................................................................................................................... 16
B. Weighting .......................................................................................................................... 17
V. Results for funds ................................................................................................................... 19
VI. Benchmark stability and alternative benchmarks ................................................................. 24
VII. Robustness of the results ...................................................................................................... 26
VIII. Performance of screened stocks ........................................................................................... 30
IX. Screening intensity and fund performance ........................................................................... 31
X. Conclusion ............................................................................................................................ 34
XI. References .......................................................................................................................... 101
Appendix A: An example of a fund’s prospectus ....................................................................... 104

3

Mutual Fund Screening Versus Weighting

Roman Skripnik
Doctor of Philosophy (BUSINESS ADMINISTRTION)
FACULTY OF THE USC GRADUATE SCHOOL
August, 2019

Abstract

This paper develops a holdings-based measure of fund performance that distinguishes how fund managers
weight stocks in their portfolios from how they screen the stocks they choose to hold. I find that screening
decisions contribute negatively to the performance of a typical fund whereas portfolio weighting decisions
contribute positively. In particular, screening decisions lower fund performance by 0.40% per year before
costs. Weighting decisions contribute 0.72% in performance per year for a typical fund during 1980-2016.
Even though the managers possess weighting ability, which in isolation suggests skill, when I also consider
the ability to pick which stocks to hold, skill is no longer present. My results also suggest that fund managers
could improve their performance by following a benchmark closer in terms of holdings but not in terms of
weights.

4

I. Introduction
Previous work on holdings-based measures of fund performance focuses on the excess returns
associated with funds’ portfolio weighting decisions. For example, in Daniel et al. (1997) and
Ferson and Mo (2016) performance is measured by the ability to put larger weights on stocks with
subsequently higher abnormal returns. A fund’s weights are often compared to benchmark weights
such as the fund’s previous holdings as in Grinblatt and Titman (1993) or a simple strategy like
buy-and-hold as in Ferson and Khang (2002). Other studies, in particular, Cremers and Petajisto
(2009) and Doshi et al. (2015) find that funds whose weights deviate from the benchmark weights
more tend to generate superior performance. While the previous literature focuses on funds’
portfolio weighting decisions, funds’ decisions in screening securities have received much less
attention. Screening is important though. The typical S&P 500 fund in my sample holds much
fewer than 500 - only 116 - stocks, which is a result of screening activity.
To understand the importance of screening decisions better, regard portfolio diversification and
screening activities as opposite to each other. Although diversification across stocks is often
considered key to portfolio construction, overdiversification carries several potential costs as well.
First, increasing the number of stocks is likely to raise monitoring costs. Because benchmarks
usually hold a considerable number of stocks and because of limited resources available to follow
all the stocks in the benchmark, funds choose not to hold many of them. For example,
Nieuwerburgh and Veldkamp (2010) argue that specialization arises because the more an investor
holds of an asset, the more valuable it is to learn about that asset, but the more an investor learns
about the asset, the more valuable that asset is to hold. In other words, specialized learning is
responsible for investors’ concentrated portfolios. Kacperczyk et al. (2016) model investment
managers with finite attention, who optimally allocate their scarce capacity to process information.
5

Both papers suggest that screening stocks out is in line with conserving manager’s attention,
resulting in lower monitoring costs.
Portfolio diversification across a large number of stocks may dilute the positive abnormal returns
from a portfolio manager’s best investment ideas. For example, Cohen et al. (2010) find that fund
managers’ best ideas not only generate statistically and economically significant risk-adjusted
returns, but also systematically outperform the rest of the positions in managers’ portfolios. The
authors argue that investors would benefit if managers held more concentrated portfolios, thus
encouraging fund managers to screen some stocks out. Finally, Shawky and Smith (2005) posit
that there is an optimal number of stocks for each mutual fund, reflecting the trade-off between
diversification benefits versus transaction and monitoring costs. They find a significant quadratic
relation between number of stock holdings and risk-adjusted returns for U.S. equity mutual fund
portfolios during 1992–2000. Shawky and Smith (2005) also find that the number of stocks held
in a fund is positively correlated with fund size, based on greater liquidity concerns for large funds.
In particular, holding a large position in one stock will make it harder to sell the positon. Managers
of smaller funds can hold more limited numbers of stocks without liquidity being a great concern.
These arguments suggest that screening is optimally balanced against diversification.
This study decomposes fund’s excess returns, calculated using stock holdings, into two terms. One
term is related to screening decisions and one is related to weighting decisions. The study finds
that funds on average do a poor job of selecting which stocks to hold. The returns from screening
range from -1.09% for S&P 500 Growth funds to 1.09% per year for Wilshire 5000, with average
losses of 0.40% per annum, before costs, across all funds.
Screening has been used in the socially responsible investing (SRI) literature, where funds apply
screens to their investments, excluding certain groups of stocks, such as gambling, alcohol or
6

tobacco companies. But screening, in various forms, is much more pervasive. Some funds
emphasize the screening process in their prospectus. Many funds use a quantitative approach to
narrow down the investment universe and screen the stocks (see Figure B1 in the appendix for an
example). Ferson et al. (1999) find that value mutual funds do not outperform growth mutual funds,
which seems inconsistent with the value anomaly that high book-to-market stocks outperform low
book-to-market ones. The authors attribute this result to the screening activity of fund managers.
My results show that screening is important to evaluating a fund’s performance, just as weighting
is. A further decomposition of the screening returns shows funds’ inability to profitably screen the
stocks in - to pick stocks that are not held in the benchmark, or to screen them out – choose not to
hold stock that are held in a fund’s benchmark. The return from screening in is negative for almost
all types of funds, with an average annual loss of 0.19% across all funds. The returns from
screening the stocks out are also negative for almost all categories of funds, with overall loss of
0.21%. The higher screening component for value funds than for growth funds – and statistically
significant difference in some cases - suggests that managers of value funds may possess better
screening ability.
This study also finds that funds do a better job weighting stocks than screening them. The return
from weighting is positive for all categories of funds, ranging from 0.16% per year for S&P
MidCap 400 to 1.46% per year for Wilshire 5000 funds. To overweight a stock, a fund manager
needs to underweight another one. My results show that the gains from the overweighted portion
are higher than the losses from the underweighted portion for each category of funds, with overall
gains of 5.16% for the overweighted portion, and overall losses of 4.44% for the underweighted
portion per year. The result that funds are able to put higher weights on stocks with (relatively)
7

higher subsequent performance and lower weights on stocks with (relatively) lower subsequent
performance is consistent with Baker et al. (2010) and Jiang et al. (2014).
Because I not only measure how well the funds allocate their portfolio weights, but also how well
they choose which stocks to hold, my results provide a more complete view of active management.
Even though the managers possess weighting ability, which in isolation suggests skill, when we
also consider the ability to pick which stocks to hold, the active component of performance,
defined as the sum of screening and weighting, is statistically insignificant, with overall annual
returns of 0.32% across all funds. My results contribute to the debate on active versus passive
funds. Since we have observed that funds lose on screening decisions, but gain on weighting
decisions, an opportunity for a fund manager to improve performance emerges by following a
benchmark closer in terms of holdings. In other words, hold more stocks from the benchmark and
fewer stocks outside the benchmark that has been assigned using Active Share measure. Regarding
weighting, managers should pursue their strategies, as they tend to beat a simple mechanical
strategy of value-weighting, consistent with Daniel et al. (1997).
The rest of the paper is organized as follows. Section II describes the data and benchmarks used,
as well as the derivation of variables used in the analysis. Section III describes the assignment of
benchmarks to funds, while section IV focuses on the calculation methodology of screening and
weighting. The main results are discussed in section V, followed by the analysis of benchmark
assignment stability in section VI. Robustness tests are presented in section VII, and the
performance analysis of screened stocks is described in section VIII. Relation between screening
intensity and fund performance is examined in section IX. Finally, section X concludes.

8

II. The data
This study examines the quarterly holdings of over 4,000 mutual funds between March 31, 1980
and December 31, 2016. I obtain this mutual fund holdings data from the Thomson Reuters Mutual
Fund Holdings file, which provides the date of funds holdings as filed with the SEC, the CUSIP
code of each holding and the number of shares held for each stock. I used the MF-Links FUNDNO-
WFICN dataset provided in WRDS by Professor Russ Wermers to obtain a unique fund identifier
(WFICN) for each fund in the Thomson Reuters holdings data. Moreover, to keep only the funds
that invest mainly in stocks, I downloaded the investment objective code (self-reported by the
fund) from the same MF-Links FUNDNO-WFICN file and retained only funds with the code 2
(aggressive growth), 3 (growth), 4 (growth and income) and 7 (balanced). I omit all the holdings
for a fund in period when the fund had fewer than 10 holdings. This is required for a meaningful
measure of Active Share, as described later.
Moreover, since the data on most of the benchmark holdings is available on a monthly frequency,
the results will be more informative if I calculate the measures of interest every month, as opposed
to every quarter. However, because the holdings’ data is available on quarterly basis – once every
three months – I assume that for the two intermediate months the number of shares held is the
same as for the reporting month, adjusted for stock splits etc. If a stock has returns data missing
for an intermediate month, this stock is omitted from that month’s holdings only, and the stock
weights are recalculated without it. Recalculating portfolio weights every month, using stock
prices at the end of every month (as opposed to every quarter), also allows for a more accurate
interpretation of results, because a more appropriate benchmark can be assigned every month,
instead of keeping it for at least 3 months. However, since some funds report only twice a year,
especially in the earlier years of the sample, the most recent holdings data may be up to six months
9

old. If the gap between two consecutive filings is more than six months, I code the holdings as
missing, following Daniel et al. (1997).
I obtain stock returns from the CRSP monthly returns file. The returns of NYSE, AMEX and
NASDAQ listed securities are used. To compute market capitalization for each firm, I download
the price and the number of shares outstanding from CRSP, along with the cumulative factors to
adjust price and shares for stock splits and dividend payouts. Finally, I multiply the price by the
number of shares for each stock held and normalize to obtain actual weights for the fund holdings.
I also use CRSP to obtain the SIC code for each stock to calculate 48 industry portfolios.
For my analysis I adjust stock returns for standard benchmark factors. To compute the benchmark
returns, I form 125 portfolios following Daniel et al. (1997). The composition of each of the 125
portfolios is based on a triple-sort on the three characteristics: size, book-to-market, and
momentum. I sort the universe of stocks at each formation date into quintiles based on market
equity just prior to the formation date, that is, on the last day of June. The breakpoints for this sort
are based on the NYSE firms only, even though the analysis includes NYSE, AMEX and
NASDAQ stocks. I further sort the firms within each size quintile into quintiles based on firms’
book-to-market ratio. The book-to-market ratio is the ratio of the book value at the end of the
firm’s fiscal year during the calendar year preceding the formation date to the market value at the
end of preceding December. I calculated book equity following Fama and French (1992) and
downloaded the necessary balance sheet components, as well as the stock exchange code (to
identify NYSE firms) for each firm from Compustat. However, I modify the Daniel et al. (1997)
approach to use a different industry normalization of book-to-market, following Wermers (2004).
In particular, I calculate the book-to-market characteristic as
ln( 𝐵𝑇𝑀 𝑖 ,𝑡 𝑗 ) −ln( 𝐵𝑇𝑀 𝑡 𝑗 )
𝜎 𝑗 [ln( 𝐵𝑇𝑀 𝑖 ,𝑡 𝑗 ) −ln( 𝐵𝑇𝑀 𝑡 𝑗 ) ]
, where 𝐵𝑇𝑀 𝑖 ,𝑡 𝑗
10

is the book-to-market ratio of stock i, which belongs to industry j on the last day of June of year t
and ln ( 𝐵𝑇𝑀 𝑡 𝑗 ) is the log book-to-market ratio of industry j (the aggregate book-value divided by
the aggregate market value). Also, 𝜎 𝑗 [ln ( 𝐵𝑇𝑀 𝑖 ,𝑡 𝑗 )− ln ( 𝐵𝑇𝑀 𝑡 𝑗 ) ] is the cross-sectional standard
deviation of the adjusted book-to-market ratio across industry j. This approach was suggested by
Cohen and Polk (1998), and used by Wermers (2004). Following Fama and French (1997), I define
48 industries depending on the underlying firm’s principal Standard Industry Classification (SIC)
code as reported by CRSP. Finally, I sort the firms in each of the 25 size/BM portfolios into
quintiles based on their preceding twelve-month return, obtaining a total of 125 portfolios.
However, the preceding twelve-month return is calculated through the end of May, that is, the
return up through one month prior to the formation date. According to Daniel et al. (1997), this
method avoids problems associated with the bid-ask bounce and monthly return reversals. Finally,
I obtained Carhart 4 factors from Kenneth French’s website, as well as the definitions of the 48
industries.
As benchmarks for the funds, I use 16 out of 19 benchmarks used by Cremers and Petajisto (2009).
I couldn’t obtain holdings for 3 indices, and thus omit them. The benchmarks used include the
S&P 500, S&P 500/Barra Growth and S&P 500/Barra Value starting December 31, 1981, S&P
MidCap 400 starting 30 June, 1991 and S&P SmallCap 600 starting 31 October, 1994. The
constituents for S&P/Barra were downloaded from Compustat and value-weighted. From the
Russell family I use Russell 1000, Russell 1000 Growth, Russell 1000 Value, Russell 2000,
Russell 2000 Growth, Russell 2000 Value, Russell 3000, Russell MidCap, Russell MidCap
Growth and Russell MidCap Value. I used ETFs to proxy for the Russell indices and used value
weights for all the holdings. All the ETFs tracking Russell indices that I found start 31 March,
2003. I also use the Wilshire 5000 index as a benchmark, which I proxy with an ETF as well, and
11

it starts 31 March, 2003. The weights for Wilshire 5000 ETF were computed using market
capitalization. I delete all the holdings of a benchmark for a month when there were fewer than 50
holdings reported, for a more accurate matching of benchmarks to funds – for the missing months
the second best matching benchmark is assigned to a fund.

III. Selection of the benchmark index
Obtaining reliable measures of fund performance requires the knowledge of actual rather than self-
reported benchmarks. For example, Sensoy (2009) shows that almost one third of actively
managed funds specify a benchmark index in their prospectus that doesn’t most closely match the
actual style of the fund. Therefore, I don’t rely on the funds’ investment objective code in CRSP
to categorize the funds. Instead, I compute the active share from Cremers and Petajisto (2009) of
a fund with respect to all 16 indices every period and assign the index with the lowest Active Share
as the fund’s benchmark for that period. Active Share is calculated the following way:
𝐴𝑐𝑡𝑖𝑣𝑒 𝑆 ℎ𝑎𝑟𝑒 =
1
2
∑|𝑤 𝑓𝑢 𝑛𝑑 ,𝑖 − 𝑤 𝑖𝑛𝑑𝑒𝑥 ,𝑖 |
𝑁 𝑖 =1
( 1)
Where 𝑤 𝑓𝑢𝑛𝑑 ,𝑖 and 𝑤 𝑖𝑛𝑑𝑒𝑥 ,𝑖 are stock i weights in the fund and in the index respectively. By
construction, the index with the lowest Active Share value has the greatest overlap with the stock
holdings of a particular fund. This approach to benchmark selection likely understates the impact
of screening, and is therefore a conservative choice for my purposes. By recalculating Active Share
every month for each fund, I can track a fund’s style changes over time. Before March 2003, when
Russell and Wilshire ETFs became available, the funds in my dataset could only have one of the
S&P indices as benchmarks. Table 1 displays some descriptive statistics for the funds and their
12

benchmarks. In particular, the average number of stocks held in a fund, its benchmark and the
average number of stocks held in common. The table shows that each category of funds holds on
average fewer stocks than its benchmark, indicating a large extent to which funds engage in stock
screening. For example, in my sample the S&P 500 holds on average 417 stocks
1
, whereas funds
that are benchmarked against S&P 500 hold on average only 116 stocks.

IV. Methodology
To understand the methodology, consider Figure 1 below, where U (the grey background) is the
universe of available stocks, F (the pink circle) are the stocks held in a fund, B (the blue circle) are
the stocks held in a benchmark, and C (the intersection of blue and pink circles) are the stocks held
in common.
Figure 1. Graphic depiction of the stocks held by the fund and its benchmark
U is the entire universe of stocks, F are the stocks held in a fund, B are the stocks held in a benchmark, and C are the
stocks held commonly by the fund and the fund’s benchmark.

1
The number of stocks is less than 500, because I omit the stocks from my dataset that had missing data on book
value, or fewer than 6 months of returns in a given year.
13

Define a fund’s before-cost returns as 𝒘 𝒕 ′
𝒓 𝒕 +𝟏 , where 𝒘 𝒕 is the portfolio weight vector. The
benchmark returns are 𝒙 𝒕 ′
𝒓 𝒕 +𝟏 , where 𝒙 𝒕 is the benchmark’s weight vector. The vector of stock
returns 𝒓 𝒕 +𝟏 = (
𝒓 𝒕 +𝟏 𝑩 𝒓 𝒕 +𝟏 𝑪 𝒓 𝒕 +𝟏 𝑵𝑩
) consists of three stacked subvectors: 𝒓 𝒕 +𝟏 𝑩 is the vector of returns for stocks
that are held in the benchmark, but not in the fund (B-C in Figure 1), 𝒓 𝒕 +𝟏 𝑵𝑩
is the vector of returns
for stocks that are held in the fund but not in the benchmark (F-C in Figure 1), and 𝒓 𝒕 +𝟏 𝑪 is the
vector of returns for stocks that are held both in the fund and the benchmark (C in Figure 1). In
other words, 𝒓 𝒕 +𝟏 is the union of stock returns held either in the fund or the benchmark (F+B-C in
Figure 1). The definition of 𝒓 𝒕 +𝟏 is therefore fund specific. Table 2 displays the value weighted
returns for each subvector in 𝒓 𝒕 +𝟏 , in addition to the returns of stocks not held by any fund. We
can see from Table 2 that the stocks held in a benchmark but not by funds perform better than the
stocks held by funds but not in a benchmark. The last row displays the difference in returns. Even
though the results are not significant, in overall the stocks held only by funds underperform those
held only in benchmarks by 0.6 percent per year. The stocks held in value funds, benchmarked
against S&P 500 Value and Russell 2000 Value, performed better on average than the stocks in
the benchmarks, although the result is insignificant. Table 3 displays the same results, but returns
are adjusted for the DGTW characteristics: size, book-to-market ratio and momentum. The results
are consistent with those in Table 2, in particular, stocks held only in the benchmark beat the stocks
held only in a fund. Although, positive values for both funds and benchmarks overall in Table 3
suggest that stocks held in both funds and benchmarks beat average stocks matched on size, book-
to-market and momentum. Moreover, the stocks held in benchmarks (but not in funds) beat
average stocks matched on these three characteristics by 0.5 percent per year, and this result is
statistically significant.
14

Another interesting result is shown in the last row of Table 2 for S&P 500 Growth and Value, and
Russell 2000 Growth and Value funds. The stocks held in the Value funds outperformed the stocks
held in the funds’ benchmarks, whereas the stocks held in Growth funds underperformed the
benchmark stocks. This result is even stronger after adjusting for DGTW characteristics. In
particular, in Table 3 not only S&P 500 and Russell 2000 funds exhibit this feature, but also Russell
1000 Growth and Value funds. This suggests that funds whose investment universe is value stocks
are better at screening than the funds that invest mainly in growth stocks, and the fact that this
result holds after adjusting for DGTW characteristics implies that it is the result of screening
ability.
Next let’s define the vector of benchmark weights, 𝒙 𝒕 = (
𝒙 𝒕 𝑩 𝒙 𝒕 𝑪 𝟎 ), which consists of three stacked
subvectors: 𝒙 𝒕 𝑩 , the vector of benchmark stock weights for stocks held in the benchmark but not
the fund, 𝒙 𝒕 𝑪 , the vector of benchmark weights for commonly held stocks, and a vector of zeros for
stocks that are held in the fund but not in the benchmark. The vector of fund weights, 𝒘 𝒕 =
(
𝟎 𝒘 𝒕 𝑪 𝒘 𝒕 𝑵𝑩
), consists of three stacked vectors as well. In 𝒘 𝒕 𝑪 I collect the fund’s weights of stocks that
are held in the fund and in the fund’s benchmark, 𝒘 𝒕 𝑵𝑩
is the vector of weights for stocks held in
the fund, but not in the benchmark. The entry is 𝟎 where a stock is held in the benchmark but not
in the fund. Both 𝒙 𝒕 and 𝒘 𝒕 sum to 1.
Value weighting provides a set of benchmark weights for stocks in the fund, and we interpret
funds’ deviation from value weights in these stocks as a weighting decision. To implement this,
let 𝒘 𝒕 𝑴 be the value-weighted vector of stocks held in the fund but not in the benchmark,
15

normalized to sum to the same number as the weights in 𝒘 𝒕 𝑵𝑩
. Let 𝒙 𝒕 𝑪 ∗
contain the same elements
as 𝒙 𝒕 𝑪 , normalized such that 𝟏 ′
𝒙 𝒕 𝑪 ∗
=1 − 𝟏 ′
𝒘 𝒕 𝑴 . We can now express fund’s excess returns over the
benchmark returns as follows:
𝒘 𝒕 ′
𝒓 𝒕 +𝟏 − 𝒙 𝒕 ′
𝒓 𝒕 +𝟏 = [(
𝟎 𝒘 𝒕 𝑪 𝒘 𝒕 𝑵𝑩
) − (
𝒙 𝒕 𝑩 𝒙 𝒕 𝑪 𝟎 )]
′
(
𝒓 𝒕 +𝟏 𝑩 𝒓 𝒕 +𝟏 𝑪 𝒓 𝒕 +𝟏 𝑵𝑩
) =
= (
𝟎 − 𝒙 𝒕 𝑩 𝒙 𝒕 𝑪 ∗
− 𝒙 𝒕 𝑪 𝒘 𝒕 𝑴 − 𝟎 )
′
(
𝒓 𝒕 +𝟏 𝑩 𝒓 𝒕 +𝟏 𝑪 𝒓 𝒕 +𝟏 𝑵𝑩
) + (
𝟎 − 𝟎 𝒘 𝒕 𝑪 − 𝒙 𝒕 𝑪 ∗
𝒘 𝒕 𝑵𝑩
− 𝒘 𝒕 𝑴 )
′
(
𝒓 𝒕 +𝟏 𝑩 𝒓 𝒕 +𝟏 𝑪 𝒓 𝒕 +𝟏 𝑵𝑩
) =
= 𝑆 + 𝑊 ( 2)
In my sample all the stocks in benchmarks are value-weighted, so (
𝟎 𝒙 𝒕 𝑪 ∗
𝒘 𝒕 𝑴 ) is a value-weighted
vector of stocks held in a fund. Equation ( 2) decomposes the excess fund returns into components
related to screening, 𝑆 , and weighting, 𝑊 . Each of the weight vectors in the 𝑆 and 𝑊 components
in equation ( 2) sum to zero, so the excess return of a fund relative to its benchmark is expressed
as the sum of an excess return due to screening and an excess return due to weighting decisions.
To illustrate equation ( 2) , assume that benchmark holds a number, B, of stocks, the fund holds F
stocks, and they commonly hold C stocks. Assume that all stocks have the same value and are
equally weighted. In this example, 𝒙 𝒕 𝑩 would sum to
𝐵 −𝐶 𝐵 , 𝒙 𝒕 𝑪 would sum to
𝐶 𝐵 , and their sum, 𝒙 𝒕 𝑩 +
𝒙 𝒕 𝑪 , is 1. Similarly, 𝒘 𝒕 𝑵𝑩
and 𝒘 𝒕 𝑴 each sum to
𝐹 −𝐶 𝐹 , and 𝒘 𝒕 𝑪 and 𝒙 𝒕 𝑪 ∗
each sum to
𝐶 𝐹 . Since these
vectors all sum to 1, it can be seen that the screening component is a long-short portfolio, where
the long portfolio is the value weighted portfolio of actual fund holdings, and the short portfolio is
the value weighted portfolio of the stocks held in the fund’s benchmark. Likewise, the weighting
16

portfolio is a long-short portfolio of stocks, where the long portfolio is the one actually held by the
fund, and the short portfolio is the portfolio consisting of the same stocks, but value-weighted.
Adding these two portfolios together results in the excess returns of the fund over its benchmark.

A. Screening
Screening ability reflects how well the fund manager is able to screen stocks in (choose to hold
stocks that are not in the benchmark) and also how well she screens them out (chooses not to hold
stocks that are held in the benchmark). One can see from equation ( 2) that if a fund holds all the
stocks value-weighted, then 𝑊 , the weighting component, is zero, implying that the fund engages
in screening only. A screener earns the returns of the stocks screened in and gives up the returns
of the stocks screened out.
From Table 1 we see that funds hold fewer stocks on average than the benchmarks. For example,
in my sample the S&P 500 holds on average 417 stocks that meet my data requirements, whereas
funds that are benchmarked against S&P 500 hold on average only 116 stocks, of which only 88
are held in common. If we assume that stocks both in funds and benchmarks are equally weighted,
then the weights of stocks held in common add up to a much bigger number for the fund than the
benchmark, because of the extensive screening out. For example, in the case of the S&P 500, the
stock weights held in common sum to 88/116=0.76 for a fund, whereas they sum to 88/417=0.21
for the benchmark. As a result, we cannot just multiply these weights by the subsequent returns
for the stocks that are held in a benchmark but not in the fund, and compare them to the weighted
returns of stocks held in a fund, but not in the benchmark.
17

To separate the screen-in and -out decisions, introduce a vector 𝒙̃
𝒕 𝑪 , which sums to 1, such that
𝒙̃
𝒕 𝑪 =
𝒙 𝒕 𝑪 𝟏 ′
𝒙 𝒕 𝑪 =
𝒙 𝒕 𝑪 ∗
𝟏 ′
𝒙 𝒕 𝑪 ∗
. With this vector I break down the screening term as:
𝑆 = (
𝟎 − 𝒙 𝒕 𝑩 𝒙 𝒕 𝑪 ∗
− 𝒙 𝒕 𝑪 𝒘 𝒕 𝑴 − 𝟎 )
′
(
𝒓 𝒕 +𝟏 𝑩 𝒓 𝒕 +𝟏 𝑪 𝒓 𝒕 +𝟏 𝑵𝑩
) = [(
𝟎 𝒙̃
𝒕 𝑪 𝟎 ) − (
𝒙 𝒕 𝑩 𝒙 𝒕 𝑪 𝟎 )]
′
(
𝒓 𝒕 +𝟏 𝑩 𝒓 𝒕 +𝟏 𝑪 𝒓 𝒕 +𝟏 𝑵𝑩
) + [(
𝟎 𝒙 𝒕 𝑪 ∗
𝒘 𝒕 𝑴 ) − (
𝟎 𝒙̃
𝒕 𝑪 𝟎 )]
′
(
𝒓 𝒕 +𝟏 𝑩 𝒓 𝒕 +𝟏 𝑪 𝒓 𝒕 +𝟏 𝑵𝑩
) ( 3)
The first, screen-out, component of (3) compares the performance of stocks held in the benchmark
only (B-C in Figure 1) to the commonly held stocks (C in Figure 1). The second, screen-in,
component of equation (3) compares the performance of stocks held in the fund only (F-C in Figure
1) to the commonly held stocks (C in Figure 1). Table 4 shows how stocks that funds screened-in
performed compared to those that funds screened-out
2
. Overall, the screened-in stocks
underperform the screened-out ones by 0.6 percent per year. On a fund category level the results
are similar for most but few categories of funds, suggesting that the screening component will be
negative.

B. Weighting
The weighting component of equation (2) captures how well the fund manager is able to weight
the stocks he holds, as compared to benchmark weights. Since the stocks in the benchmarks that I
use are value-weighted, the 𝒙 𝒕 𝑪 component is value-weighted. The weighting component is
therefore similar to the active weight measure in Doshi et al. (2015). Their measure is the absolute
difference between the value weights and actual weights held by a fund, summed across its

2
The “screen-out” and “screen-in” rows in Table 4 corresponds exactly to the “benchmark only” and “fund only”
rows in Table 2.
18

holdings. The weighting measure, W, is zero if the fund holds all the stocks held in the benchmark
using benchmark weights, and if it uses value-weights for stocks not held in the benchmark. I
further decompose the weighting measure into returns from overweighting and returns from
underweighting stocks:
𝑊 = {[𝑰 𝑤 𝑡 𝐶 ≥ 𝑤 𝑡 𝑀 ∗ ( 𝒘 𝒕 𝑪 − 𝒙 𝒕 𝑪 ∗
) ]
′
𝒓 𝒕 +𝟏 𝑪 + [𝑰 𝑤 𝑡 𝑁𝐵
≥ 𝑤 𝑡 𝑀 ∗ ( 𝒘 𝒕 𝑵𝑩
− 𝒘 𝒕 𝑴 ) ]
′
𝒓 𝒕 +𝟏 𝑵𝑩
}
− {− [𝑰 𝑤 𝑡 𝐶 <𝑤 𝑡 𝑀 ∗ ( 𝒘 𝒕 𝑪 − 𝒙 𝒕 𝑪 ∗
) ]
′
𝒓 𝒕 +𝟏 𝑪 − [𝑰 𝑤 𝑡 𝑁𝐵
<𝑤 𝑡 𝑀 ∗ ( 𝒘 𝒕 𝑵𝑩
− 𝒘 𝒕 𝑴 ) ]
′
𝒓 𝒕 +𝟏 𝑵𝑩
} ( 4)
In the equation ( 4) the part in the first curly brackets shows how much the fund manager gains
from overweighting the stocks, and the part in the second curly brackets shows how much the
manager gains form underweighting stocks. The second component is with a negative sign because
underweighted stocks will have a negative net weight, 𝒘 𝒕 𝑪 − 𝒙 𝒕 𝑪 ∗
, thus the portfolio return, in
expectation, will be negative, thus, to have a positive return for an easier interpretation of results,
we have to multiply returns from underweighting by negative one. In this case both underweighting
and overweighting measures will be positive, and thus we only compare the magnitudes of both to
determine the manager’s weighting skill. One can also see that, when the overweighting
component is calculated, only the stocks whose weights are bigger than value weights are
considered (likewise for underweighting). As a result, the difference between the overweighed
weights and value weights is always positive. Since the stock returns are positive in expectation,
multiplying vector of positive weight differences by the vector of stock returns results in a big
number. Table 4 shows that the stocks that funds overweight overall beat the stocks that funds
underweight by 1.18 percent a year, and this result is positive for all the fund styles, except those
benchmarked against Russell 1000 and Russell 1000 Growth. The results stay consistent even after
DGTW adjustment of stocks returns. The stocks that funds overweighted beat those that they
19

underweighted by 0.4 percent per year, with negative results for the same Russell 1000 and Russell
1000 Growth funds, as well as S&P SmallCap 600. This suggests that the funds are able to put
higher weights on stocks that subsequently perform better than those with lower weights,
predicting positive weighting component in equation ( 2) .

V. Results for funds
The previous tables examined the performance of stocks, grouped according to how funds treat
them. In this section I compute hypothetical fund returns as the returns of the portfolio of NYSE,
AMEX and NASDAQ listed equities reported in the funds’ quarterly reports. The portfolio weights
are multiplied by the subsequent monthly returns of securities and summed to construct
hypothetical monthly returns for each fund. Benchmark returns are calculated the same way, using
value-weights. Table 6 provides the decomposition of fund performance according to their
screening and weighting decisions. I compute the measures for each fund in every month and
value-weight them across funds in a group for each month and then average these over time. I
separate funds into groups according to the benchmark for each fund. The screening component is
further broken down into the returns from holding stocks that are not held in the benchmark
(screen-in) and from not holding stocks that are in the benchmark (screen-out). The weighting
component is broken down into the returns from overweighting stocks and from underweighting
them. Each column represents the benchmark used for calculating the measures. The number of
funds is significantly greater than the initial 4,000, because I allow funds’ benchmarks to change
from month to month, and, each time a fund’s benchmark changes, it is considered a different
fund. This means that if a fund belonged to S&P 500 category for 20 months, and to the Russell
20

1000 for the next 10 months, its returns contributed to the results of S&P 500 column in Table 6
for 20 months and to the results of Russell 1000 for the next 10 months, and it would be counted
as 2 funds instead of just 1, even though it is the same fund.
The screening component shows that funds on average do a poor job of selecting which stocks to
hold. Although insignificant, the screening measure is negative overall and for almost all of the
fund styles, with average losses of 0.4% per annum attributed to screening. Moreover, as suggested
by the results in Table 2, the screening component is mostly positive for value funds and noticeably
higher than the screening component for growth funds, supporting the previous finding that
managers of value funds possess better screening ability. On the other hand, funds do a much better
job weighting stocks, in particular, overweighting or underweighting stocks as compared to value-
weighting. Despite its statistical insignificance, the weighting component is mainly positive, with
an average of 0.72% per year across all funds. The active component, sum of screening and
weighting, is positive on average, with 0.32% return per year, although insignificant. The results
show that, although the screening component has been overlooked in the previous literature, it is
an important component of the performance of the fund. In particular, funds on average gain from
weighting, but they lose from screening.
Decomposing the screening measure further, both screen-out and screen-in components are
negative overall, suggesting that fund managers screen-out stocks that performed better than the
ones commonly held, and screened-in stocks that performed worse than the commonly held ones.
At the fund style level, the results are mixed, in particular, for most styles either screen-out or
screen-in is positive and another one is negative, resulting in varying signs for the screening ability
across various fund styles. However, funds benchmarked against S&P SmallCap 600, Russell 1000
Growth and Russell MidCap Growth saw both screen-in and screen-out returns negative. Funds
21

benchmarked against Russell 1000 Growth experienced the lowest active returns (excess returns
over the benchmark returns) across all the fund styles of -0.63 percent per year. On the other hand,
funds benchmarked against Russell 3000 and Wilshire 5000 had both screen-in and screen-out
positive. Wilshire 5000 funds also experienced the highest active returns across all the fund styles
of 2.56 percent per year. Results in Table 6 also show that the positive screening component for
value funds occurs because of positive screen-out component, which is statistically significant for
the S&P 500 Value funds. Although there are a few exceptions, screen-out, screen-in and screening
components are mainly insignificant.
Furthermore, the decomposition of weighting measure shows the gains form overweighting and
the losses from underweighting stocks. In particular, since all the weights have to sum to 1,
overweighting some stocks means that fund manager has to underweight some other ones, and the
portion of weights above the value-weights for the stocks that have been overweighted has to
match the portion of weights below the value-weights for the underweighted stocks, so that all the
weights still sum to 1. So, the overweight component in Table 6 shows the gains of that portion of
the weights above the value weights, and underweight shows the losses of that portion of weights
below the value weights. When you subtract underweight component from overweight, you get
the excess returns of portfolio over the same, but value-weighted, portfolio, i.e. the weighting
component. Table 6 shows that the gains from overweighted portion of weights is higher than the
losses from underweighted portion of weights for each category of funds, with overall gains of
5.16 percent for overweighted portion, and overall losses of 4.44 percent for underweighted
portion per year, making the weighting component positive for all the categories of funds, and
overall gains of 0.72 percent per year. These results show that funds are able to put higher weights
on stocks with subsequent high performance and underweight stocks with subsequent low
22

performance. The result is consistent with Baker et al. (2010), where the authors find that for an
average fund the subsequent earnings announcement returns on the fund’s weight-increasing
stocks are significantly higher than those on the weight-decreasing stocks. Moreover, Jiang et al.
(2014) compute the difference between the stock’s weight in each individual mutual fund portfolio
and its weight in the stock index against which that fund is benchmarked, and call this measure
deviation from benchmark (DFB). They find that stocks in the decile portfolio with the highest
DFB, which are those most heavily overweighted by mutual funds, perform substantially better
than those with the lowest DFB, which is consistent with my findings.
In Table 7, I adjust the vector of stock returns by subtracting the portfolio returns of one of the
DGTW 125 characteristic-based benchmarks for that stock. The screening component remains
negative overall, with positive values for some of the fund styles. However, the screen-in
component becomes positive, suggesting that after adjusting for stock characteristics, the stock
that managers screen in outperform those commonly held between the funds and its benchmark.
The negative overall screening measure, though, shows that the stocks that managers screened in
still underperform the stocks they screened out. The weighting and the active component maintain
their signs overall, suggesting robustness of the results overall. However, at the style level some
funds lose more from the underweighted portion of weights than they gain from the overweighted
portion after adjustment for stock characteristics. But even more interesting is that S&P 500, S&P
500 Growth, Russell MidCap Growth and Wilshire 5000 categories of funds experience negative
underweight component, which means that they gain form the underweighted portion of the
weights. This shows that some categories of funds are very good at weighting stocks, even after
adjusting for stock characteristics. Consistent with Table 6, the results in Table 7 suggest that funds
23

lose at picking which stocks to hold, but gain at weighting them and with a higher magnitude,
resulting in positive active component, thus, active portfolio management helps fund performance.
Reconciling the evidence presented so far, we can contribute to the debate on active versus passive
funds. However obvious it may sound, but there is always optimal level of managerial activeness
– my results suggest where that optimal level is. Since we have observed that funds lose on
screening decisions, but gain on weighting decisions, an opportunity for a fund manager to
improve the performance emerges by following a more passive strategy: since we have observed
that funds lose on screening decisions, but gain on weighting decisions, an opportunity for a fund
manager to improve the performance emerges by following a benchmark closer in terms of
holdings but not the weights. In other words, hold more stocks from the benchmark and fewer
stocks outside the benchmark that has been assigned using Active Share measure. Regarding
weighting, managers should pursue their strategies, as they tend to beat a simple mechanical
strategy of value-weighting. My findings complement the results of Cremers and Petajisto (2009),
who show that funds that deviate more from their benchmarks’ holdings tend to obtain higher
alphas. The authors, however, do not distinguish between the contribution of stock selection and
weighting to their results. I suggest that the deviation of portfolio weights from the benchmark
weights, and not the deviation of holdings from the benchmark holdings, is the main driver of these
higher alphas.
Table 8 displays results for a different definition of overweighting and underweighting, as well as
a new measure that I call evenweight. Here over and underweighted weights are defined not just
as weights that are above or below value weights, but above or below by some nonzero number.
In particular, I set this number to be
1
# 𝑜𝑓 ℎ𝑜𝑙𝑑𝑖𝑛𝑔𝑠 , or equal weight. This results in the following
24

definitions of measures: 𝑜𝑣𝑒𝑟𝑤𝑒𝑖𝑔 ℎ𝑡 ≥ 𝑣𝑎𝑙𝑢𝑒 𝑤𝑒𝑖𝑔 ℎ𝑡 +
1
# 𝑜𝑓 ℎ𝑜𝑙𝑑𝑖𝑛𝑔𝑠 ; 𝑢𝑛𝑑𝑒𝑟𝑤𝑒𝑖𝑔 ℎ𝑡 ≤
𝑣𝑎𝑙𝑢𝑒 𝑤𝑒𝑖𝑔 ℎ𝑡 −
1
# 𝑜𝑓 ℎ𝑜𝑙𝑑𝑖𝑛𝑔𝑠 .
The weights that fall in between overweight and underweight, are called evenweights and show
how stocks, whose weights don’t deviate much from value weights, perform: 𝑣𝑎𝑙𝑢𝑒 𝑤𝑒𝑖𝑔 ℎ𝑡 +

1
# 𝑜𝑓 ℎ𝑜𝑙𝑑𝑖𝑛𝑔𝑠 > 𝑒𝑣𝑒𝑛𝑤𝑒𝑖𝑔 ℎ𝑡 > 𝑣𝑎𝑙𝑢𝑒 𝑤𝑒𝑖𝑔 ℎ𝑡 −
1
# 𝑜 𝑓 ℎ𝑜𝑙𝑑𝑖𝑛𝑔𝑠 . In most of the cases overweighted
returns are lower than underweighted, but the weighting component remains positive because the
new measure, evenweight, is always positive. This suggests that most heavily overweighted stocks
underperform the most heavily underweighted ones, with the stocks in between significantly
contributing towards positive weighting returns. Table 9 displays the results with evenweight, but
adjusted for DGTW portfolio returns. After adjusting the returns for stock characteristics,
overweighted stocks outperform the underweighted ones, while the stocks in between contribute
to positive weighting returns as well.

VI. Benchmark stability and alternative benchmarks
I treat a benchmark as known to the manager, relative to which he makes screening decisions. In
particular, each manager knows what stocks he holds and their weights. Assuming he has
knowledge of the benchmarks and their weights, the manager also knows what benchmark would
be the closest to the fund’s holdings. Moreover, before deciding which stocks to buy or sell, I
assume the manager also understands what would be the most suitable benchmark after the
reconstitution of the portfolio. Since my dataset consists of 4,157 unique funds, and there are
16,176 fund-style assignments, a typical fund changed its benchmark on average 3 times
25

throughout the 1980-2016 period. Moreover, since an average fund in my dataset exists for 145
months (12 years), each benchmark assignment lasts on average for 3 years. Table 10 displays a
transition matrix for the benchmark assignments, i.e., it shows what is the probability of assigning
a particular benchmark at t+1, given a benchmark at t. For example, the first row shows that if a
fund had the S&P 500 as a benchmark in period t, there is 75.2% chance that the fund will be
assigned to S&P 500 in t+1, or there is 1.6% chance that the fund will be assigned to Russell 1000
in t+1. My method of benchmark selection is no doubt noisy. To see how an error in benchmark
selection might affect my results, I introduce alternative sets of benchmarks and analyze how
applying them impacts the measures of interest.
Since I used ETFs to proxy for the Russell indices, the holdings data for Russell indices starts in
2003, whereas funds’ holdings data starts in 1980. To expand the lengths of the time-series over
which I have the data on Russell holdings, I replicate the method used by the Frank Russell
company (FTSE Russell, Construction and Methodology, Russell U.S. Equity Indexes, v3.5) to
construct portfolios all the way back to 1980. Table 11 displays the results. Russell 1000 changed
the sign of screening component, but still remains insignificant. Russell 2000 remains positive, but
becomes highly significant. Russell MidCap Value became positive, whereas Wilshire 5000
became negative. Overall, screening became less negative, and screen-in is positive.
I also formed 48 industry portfolios following Fama and French (1997), value weighted them,
added them to expand the list of benchmarks and repeated the analysis. Table 12 displays the
results. After controlling for funds tracking industry portfolios, screening became less negative as
well; for some categories of funds screening became positive and more significant. The results in
the second and the third columns in Table 12 show that funds benchmarked against a broader
benchmark have negative screening component, like before, whereas more specialized funds
26

perform much better in terms of screening, with a positive, although insignificant, screening
measure.
Now, that I obtained additional results using alternative sets of benchmarks, I can compare the
accuracy of benchmark assignments using each set. To do this I use active share measure, in
particular, I calculate average active share measure for each category of funds using each set of
benchmarks out of three. I report results in Table 13, but only for the 16 original benchmarks, i.e.
for the set of benchmarks with added 48 industry portfolios I leave out the active share values for
the industry portfolios. The results show that the benchmark assignment quality for the original
benchmarks and those obtained using Frank Russell methodology is very similar both in terms of
magnitudes and standard deviations of active share measures, with average of 0.81 and 0.82
respectively and standard deviation of 0.17 for both sets of benchmarks. However, when I add the
48 industry portfolios, the average active share measure decreases to 0.74, showing a greater
overlap of fund holdings with those of the benchmark, than with the other two sets of benchmarks.
The results suggest that the funds that were assigned to the industry portfolios as benchmarks were
the ones that had higher active share and thus lower overlap with the 16 original benchmarks. In
other words, the industry benchmarks absorbed mainly the funds that had low overlap with S&P
or Russell benchmarks, thus improving the quality and reducing the noise of benchmark selection.

VII. Robustness of the results
Currently, most of the tests use the monthly time series of portfolio returns across funds, and stocks
are double-counted in some of these portfolios. The way that the stock returns are weighted in this
calculation may not produce the most efficient estimate of the different treatment effects. Using
27

panel methods may address this issue, as well as enhance the power of my tests. Consider a panel
regression of stock returns on the left hand side, with measures of screening “treatment” on the
right hand side, along with controls. The controls, by explaining some of the variance, may
improve the precision of the estimates of screening effects. With future returns on the left hand
side, the question is whether screening is predictive of future returns.
I run a panel regression for each fund separately. The dependent variable is the future stock return,
the independent variable is the lagged screening weights, and I also add stock fixed effects and
adjust for Stambaugh (1999) bias following Hjalmarsson (2010). I remove the top and bottom
2.5% of the funds’ resulting beta coefficients (to remove the outliers) and report the results in
Table 14. The table shows average betas and t-statistics for point estimates of those betas. The
results from the panel regression are not very different from the original results, in particular, the
beta coefficients of screening weights are of a similar magnitude and sign as before, however,
adjusting for autocorrelation of the screening weights using Hjalmarsson (2010), did change the
sign of screening in a few instances. In particular, the overall screening, as well as S&P SmallCap
600 and a few Russell funds changed the sign from negative to positive. The beta coefficients for
Russell 2000 and Russell 3000 funds are unusually large, because of the small number of funds
belonging to these categories. However, more importantly, the significance of the resulting betas
remains low. The second, third and fourth columns in Table 14 show the percentage of t-statistics
falling into the [-1.65;1.65], [-1.96;1.96] and [-2.58;2.58] ranges respectively. For example,
91.91% of t-statistics for overall results (1 column) fall into the [-1.96;1.96] range, meaning that
only 8.09% of the beta coefficients are significant at 95% level. Since most of the funds have this
value above 90%, it means that statistical significance of screening is low. The results suggest that
on a fund level screening weights is not a strong predictor of subsequent stock returns.
28

I repeat the same panel regression analysis, but replace the stock returns by DGTW adjusted stock
returns. Results are reported in Table 15, and they are consistent with those in Table 14. In
particular, the magnitude of the beta coefficients is similar, and the percentage of t-statistics outside
the significance range is similar as well. One drawback, however, using Hjalmarsson (2010)
methodology is that it can’t be extended to multiple regressors, as the variable demeaning process
does not guarantee that the covariance matrix, when calculating betas, will be invertible. For this
reason, I can’t use Hjalmarsson (2010) to add controls. Instead, I run a simple panel regression
with stock fixed effects, where I add Carhart (1997) 4 factors to the right hand side as controls,
expecting these factors to explain a great portion of stock price variability. The results are reported
in Table 16, and Table 17 for DGTW adjusted stock returns. The beta coefficient of screening is
positive for almost all categories of funds, except for S&P MidCap 400 funds, both with unadjusted
and DGTW adjusted stock returns. However, in both cases the percentage of t-statistics in the
significance range is still very low.
I perform another robustness check by splitting the dataset into 2 periods. I repeat the original
analysis, but divide the entire 1980-2016 period into subperiods. Tables 18 and 20 display sample
of funds divided into 1980-1998 and 1999-2016 subperiods respectively for the benchmarks
obtained using Frank Russell benchmark construction methodology. The results clearly show that
the gross (total) returns of the funds were lower in the second half, but the screening, weighting
and thus active components were higher. This observation suggests that in periods, when markets
are doing well, the managers tend to underperform the benchmarks, with negative screening and
insignificant weighting components, however, in periods when markets are doing poorly,
managers tend to outperform the benchmarks, with positive screening and significantly positive
29

weighting components. The active component, which measures managerial skill in my setup, is
not only positive but also statistically significant in the 1999-2016 subperiod.
Tables 19 and 21 display results with DGTW adjusted stock returns for 1980-1998 and 1999-2016
subperiods respectively. The DGTW adjustment does not change my findings, in particular, that
managers underperform in good times and outperform in bad times.
In comparison, tables 22 and 24 display samples of funds divided into 1980-2002 and 2003-2016
periods respectively with the original benchmark assignments (not the ones constructed using
Frank Russell methodology). The former period only has data on the S&P family of indices,
because the data on Russell and Wilshire holdings is not available before 2003. The results are
consistent with those in tables 18 and 20, suggesting that, even though the returns for both funds
and their benchmarks were higher in the earlier subperiod, the later subperiod saw greater value
added by the fund managers. This holds true even after adjusting for DGTW returns, and the results
are displayed in Tables 23 and 25.
Finally, table 26 displays results with a few months of extreme returns removed. In particular, I
removed October of 1987, March of 2000 and September of 2008. The results without the extreme
months are, in general, consistent with the original results. In particular, screening, weighting and
active components have similar signs and magnitudes as before. However, since the overall results
became slightly less negative (-0.34 compared to -0.40 from before), I conclude that the screening
component for the three omitted months was very negative, as the three months comprise only
0.68% of the dataset (3 months out of 444), but push the screening component down from -0.34 to
-0.40. The results without the months of extreme returns, and with DGTW adjusted returns are
displayed in Table 27. Comparing the results to those in Table 7, where the months with extreme
returns are included, I find that when adjusted for DGTW returns, months with extreme returns do
30

not affect neither screening nor weighting components, and thus I can conclude that in the months
with extreme returns almost all the stocks went down, and adjusting for stock characteristics had
no effect. Combining results in Tables 18 through 27, one can conclude that fund managers do
better when markets are doing poorly, however, when markets are doing extremely poorly,
managers lose as well.

VIII. Performance of screened stocks
We have seen so far that funds screen out stocks that perform better than the stocks that funds
screen in. However, unless a fund manager follows a benchmark replication strategy, screening in
and out is inevitable. It is reasonable to assume – and the results form Table 1 confirm this – that
funds screen out stocks more often than they screen them in. Because benchmarks usually hold a
considerable number of stocks, funds choose not to hold many of them, because of transaction
costs or because of limited resources available to follow all the stocks in the benchmark. For
example, Kacperczyk et al. (2016) model investment managers with finite attention and show that
optimal attention allocation on funds’ holdings varies with the state of the economy. Because
screening in and out is inevitable, we can look at the performance of stocks based on the number
of times it has been screened in or out, in order to have a more complete view on the screening
ability. Table 28 shows the returns of value-weighed quintile portfolios of stocks sorted based on
the number of times a stock has been screened out by each of the funds in the style group and
overall.
Table 28 shows that stocks that have been screened-out more often, perform better overall than
those that have been screened-out less often, suggesting that funds often tend to not hold stocks
31

that outperform, consistent with a negative screening component. The last row of the table shows
the return of long-short portfolios, buying stocks that were screened out the most and selling stocks
that have been screen out the least. The last row shows that overall the most screened out stocks
outperform than least screened out ones by 0.69 percent per year. These results are consistent even
after adjusting the stock returns following Daniel et al. (1997), and the results are presented in
Table 29.
On the other hand, Table 30 shows the returns of value-weighed quintile portfolios of stocks sorted
based on the number of times a stock has been screened in by each of the fund style and overall.
Table 30 shows that stocks that have been screened-in more often, perform worse overall than
those that have been screened-in less often, suggesting that funds often tend to hold stocks that
underperform, which again suggests negative screening component. The last row shows that
overall the most screened in stocks underperform the least screened out ones by 0.16 percent per
year. Tables 28 and 30 together suggest that funds tend to hold stocks that perform worse than the
stocks they hold less often. Also, the stocks funds choose to screen out more often outperform the
stocks that funds screen out less often, further lowering the returns from screening. .

IX. Screening intensity and fund performance
A number of papers in the finance literature examine the link between fund performance and
socially responsible investing (SRI), which applies such screens as tobacco, alcohol, community,
employee relations, environment, and diversity. For example, Barnett and Salomon (2006)
measure how screening intensity affects the financial performance of the SRI funds. They find a
curvilinear relationship between screening intensity and financial performance - as the number of
32

social screens used by an SRI fund increases, financial returns decline at first, but then rebound as
the number of screens reaches a maximum. Humphrey and Lee (2011) investigate the relation
between screening intensity and fund performance among Australian SRI funds, finding that
screening intensity does not impact a fund’s total return, but that funds with more screens provide
weakly better risk-adjusted performance. Capelle-Blancard and Monjon (2012) examine whether
the financial performance of French SRI mutual funds is related to the characteristics and intensity
of the screening process. They find that the financial performance of SRI funds is hurt by the
exclusion of non-socially responsible stocks. Like Barnett and Salomon (2006), they also find that
this initial negative effect is partly offset as the number of screens increases.
Motivated by the findings that the returns of socially responsible funds are related to the stock
screening intensity, I analyze the entire dataset of funds in light of this relationship. In particular,
I construct the measure of screening intensity and investigate to what extent this measure predicts
fund returns. The studies mentioned in the previous paragraph proxy the screening intensity with
the number of screens the socially responsible fund applies. In particular, they either regress the
fund returns on the number of screens used (controlling for other factors), or split funds into groups
based on the number screens applied by the fund, and then analyze each group. I propose another
measure of screening intensity, which allows me to calculate this measure for any fund. The Figure
1 in methodology section will help significantly in understanding the measures (recall the
meanings of F, B and C). The proposed measures are:
𝑆𝑐𝑟𝑒𝑒𝑛 − 𝑖𝑛 𝑖𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 =
# 𝑜𝑓 𝑠𝑡𝑜𝑐𝑘𝑠 𝑖𝑛 ( 𝐹 − 𝐶 )
# 𝑜𝑓 𝑠𝑡𝑜𝑐𝑘𝑠 𝑖𝑛 𝐵 ( 6)
𝑆𝑐𝑟𝑒𝑒𝑛 − 𝑜𝑢𝑡 𝑖𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 =
# 𝑜𝑓 𝑠𝑡𝑜𝑐𝑘𝑠 𝑖𝑛 ( 𝐵 − 𝐶 )
# 𝑜 𝑓 𝑠𝑡𝑜𝑐𝑘𝑠 𝑖𝑛 𝐵 ( 7)
33

𝑆𝑐𝑟𝑒𝑒𝑛𝑖𝑛𝑔 𝑖𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 =
# 𝑜𝑓 𝑠𝑡𝑜𝑐𝑘𝑠 𝑖𝑛 ( 𝐵 + 𝐹 − 2𝐶 )
# 𝑜𝑓 𝑠𝑡𝑜𝑐𝑘𝑠 𝑖𝑛 𝐵 ( 8)
From the above equations we see that the screen-in intensity is the ratio of the number of stocks
held only in the fund to the number of stocks in the benchmark (this acts as an adjustment for the
fund style). Screen-out intensity is the ratio of the number of stocks held only in the benchmark to
the number of stocks in the benchmark. Finally, the screening intensity is the sum of the two. Table
32 shows the screening and weighting measures for the top and bottom deciles of funds, based on
their screening intensity, as well as the top-minus-bottom value. The screening intensity does not
seem to affect the screening component overall. For the funds benchmarked against Russell 1000
and Russell 2000 indices top decile performed significantly worse than the bottom decile (at 10%
level), suggesting that for these categories of funds screening intensity is associated with low
screening returns. However, for all other funds the results are insignificant. Table 33 displays the
results with DGTW adjusted stock returns. Findings are consistent with those in Table 32, but Top-
Bottom results are insignificant for all fund categories. The overall results suggest that screening
intensity is a weak predictor of the screening component.
I continue by repeating the screening intensity analysis, but instead of using just the number of
stocks to calculate the intensity measure, I weight the stocks by their market capitalization,
adjusting for the size of the screened-in or screened-out stocks. I report the results in Table 34, and
DGTW adjusted results in Table 35. The long-short portfolio, that buys top decile of fund sorted
on screening intensity and shorts the bottom decile of funds, yields insignificant results. In
particular, the return of the long-short portfolio is only significant for the Russell 1000 funds (same
with DGTW adjusted stock returns). These results are consistent with the previous results, that
screening intensity is a weak predictor of the subsequent screening return. Moreover, weighting
34

stocks by their market capitalizations when calculating screening intensity allows to see that
screening-in or screening-out a large stock doesn’t seem to be more informative that doing it with
a small stock.

X. Conclusion
To conclude, choosing which stocks to hold, as well as assigning weights for each stock are a part
of a fund decision maker’s managerial ability. Motivated by previous findings that funds’
deviations from benchmark weights are important indicators of performance, this study
decomposes the hypothetical excess returns, into returns from the screening and the weighting
decisions. This is the first study to my knowledge to focus on screening decisions using a broad
sample of funds. I find that funds on average do a poor job selecting which stocks to hold. The
results show that, although the screening component has been overlooked in the previous literature,
it is important. In particular, when considering the screening return, fund performance is negative.
On the other hand, funds do a better job weighting stocks. My findings regarding active fund
management provide a broader view relative to the existing knowledge, because I not only measure
how well the funds allocate their portfolio weights, but also how well they choose which stocks to
hold. Finally, since we have observed that funds lose on screening decisions, but gain on weighting
decisions, an opportunity for a fund manager to improve the performance emerges by following a
benchmark closer in terms of holdings but not the weights. In other words, hold more stocks from
the benchmark and fewer stocks outside the benchmark that has been assigned using Active Share
measure. Regarding weighting, managers should pursue their strategies, as they tend to beat a
simple mechanical strategy of value-weighting, consistent with Daniel et al. (1997).
35

Table 1. Average number of stocks in a fund, in its benchmark and commonly held over 1980-2016 for each category of funds
The first row shows how many stocks on average a fund holds (F in Figure 1), based on its style. The second row shows the average number of holdings for each
benchmark (B in Figure 1), and the last row shows how many stocks, on average, are held in common by the fund and its benchmark (C in Figure 1).
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Average number of
stocks in a fund
191 116 37 35 103 121 258 54 55
Average number of
stocks in a benchmark
567 417 124 241 347 524 715 476 448
Average number of
common stocks
136 88 13 15 39 58 197 34 36

Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Average number of
stocks in a fund
926 167 96 62 217 47 1094 416
Average number of
stocks in a benchmark
1228 904 774 353 586 393 1835 1415
Average number of
common stocks
698 106 58 36 123 25 965 306

36

Table 2. Average yearly gross returns (%) of portfolios formed using stocks held only in the fund, only in the benchmark,
commonly held or not held at all, grouped by fund’s benchmark
For each of the 4 groups of stocks (U is the universe of available stocks, F are the stocks held in a fund, B are the stocks held in a benchmark, and C are the stocks
held commonly by the fund and the fund’s benchmark) I calculate value weighted portfolio returns for each fund. Then I calculate averages for each of the 4 groups
of stock portfolios across all the funds in a specific style category each month. Finally, I calculate the time-series average for each fund style and report the results.
T-statistics are calculated using standard errors of time series mean and are presented in brackets.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 16176 2402 2869 2229 1395 932 907 993 1146

Fund only (F-C)
12.35 12.21 12.65 15.11 11.95 11.41 7.99 9.76 7.75
(4.52) (4.40) (3.89) (5.02) (3.37) (2.84) (1.50) (2.12) (1.48)
Benchmark only (B-C)
12.95 12.09 14.67 13.19 12.77 11.93 9.67 10.00 10.32
(5.18) (4.84) (4.88) (4.50) (3.92) (3.00) (2.05) (2.23) (2.32)
Commonly held (C)
12.69 11.81 15.31 12.26 14.49 12.05 11.45 8.94 10.12
(4.87) (4.65) (4.27) (4.16) (3.93) (3.02) (2.59) (2.02) (2.29)
Not held (U-F-B+C)
11.92 12.08 13.19 13.85 10.21 10.20 7.26 8.49 7.81
(4.58) (4.21) (4.48) (4.50) (3.53) (3.15) (1.33) (1.68) (1.52)
Fund only – Benchmark
only (F-B)
-0.60 0.12 -2.03 1.91 -0.82 -0.52 -1.68 -0.24 -2.57
(-0.92) (0.09) (-1.38) (1.78) (-0.73) (-0.41) (-0.77) (-0.15) (-1.51)

37

Table 2 (continued). Average yearly gross returns (%) of portfolios formed using stocks held only in the fund, only in the
benchmark, commonly held or not held at all, grouped by fund’s benchmark
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 187 264 316 479 199 305 315 1238

Fund only (F-C)
12.19 12.24 10.40 9.25 7.87 10.27 10.18 12.38
(2.24) (2.15) (1.74) (1.73) (1.39) (1.84) (1.74) (2.40)
Benchmark only (B-C)
11.98 11.46 11.39 9.86 10.52 10.35 10.11 13.04
(2.08) (1.95) (1.89) (1.83) (1.94) (1.91) (1.83) (3.20)
Commonly held (C)
12.67 11.20 10.62 9.56 9.64 10.00 10.77 13.07
(2.22) (1.92) (1.79) (1.78) (1.76) (1.77) (2.36) (3.22)
Not held (U-F-B+C)
10.59 9.45 9.12 8.65 9.05 8.80 7.81 11.89
(2.37) (2.01) (1.93) (1.86) (1.92) (1.85) (1.42) (2.57)
Fund only –
Benchmark only (F-B)
0.20 0.78 -0.99 -0.60 -2.64 -0.08 0.07 -0.66
(0.09) (0.41) (-0.62) (-0.47) (-1.66) (-0.04) (0.03) (-0.31)

38

Table 3. Average yearly excess (DGTW) returns (%) of portfolios formed using stocks held only in the fund, only in the
benchmark, commonly held or not held at all, grouped by fund’s benchmark
For each of the 4 groups of stocks (U is the universe of available stocks, F are the stocks held in a fund, B are the stocks held in a benchmark, and C are the stocks
held commonly by the fund and the fund’s benchmark) I calculate value weighted portfolio returns for each fund, adjusting the vector of stock returns following
Daniel et al. (1997). Then I calculate averages for each of the 4 groups of stock portfolios across all the funds in a specific style category each month. Finally, I
calculate the time-series average for each fund style and report the results. T-statistics are calculated using standard errors of time series mean and are presented in
brackets.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 16176 2402 2869 2229 1395 932 907 993 1146

Fund only (F-C)
0.21 0.83 -0.57 1.23 0.75 0.46 -1.01 1.01 -0.91
(0.56) (1.05) (-0.87) (2.12) (1.14) (0.77) (-0.62) (1.44) (-1.00)
Benchmark only (B-C)
0.53 0.13 1.24 -0.29 0.83 0.30 0.35 0.66 1.01
(2.81) (0.54) (2.05) (-0.53) (1.23) (0.38) (0.64) (1.18) (1.64)
Commonly held (C)
0.55 0.15 1.64 -0.38 2.25 0.51 2.16 0.08 1.28
(1.44) (0.38) (1.84) (-0.45) (2.45) (0.53) (2.76) (0.08) (1.20)
Not held (U-F-B+C)
-0.40 -0.74 -0.46 0.17 -0.07 -0.02 -1.77 -0.47 -1.07
(-2.62) (-2.08) (-1.38) (0.60) (-1.39) (-0.75) (-2.25) (-0.95) (-2.41)
Fund only – Benchmark
only (F-B)
-0.33 0.69 -1.82 1.51 -0.08 0.16 -1.36 0.35 -1.92
(-0.75) (0.80) (-1.83) (1.85) (-0.08) (0.17) (-0.75) (0.32) (-1.57)

39

Table 3 (continued). Average yearly excess (DGTW) returns (%) of portfolios formed using stocks held only in the fund, only in
the benchmark, commonly held or not held at all, grouped by fund’s benchmark
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 187 264 316 479 199 305 315 1238

Fund only (F-C)
0.65 1.61 0.18 0.40 -0.89 0.53 -0.56 -0.86
(0.53) (2.22) (0.20) (0.40) (-0.66) (0.27) (-0.25) (-0.50)
Benchmark only (B-C)
-0.40 0.42 1.28 0.08 0.22 0.39 -0.95 -0.06
(-0.33) (0.39) (1.36) (0.09) (0.30) (0.52) (-1.00) (-0.14)
Commonly held (C)
-0.37 -0.03 0.84 0.39 -0.46 0.17 0.21 0.11
(-0.33) (-0.03) (0.56) (0.30) (-0.35) (0.09) (0.18) (0.06)
Not held (U-F-B+C)
-0.08 -0.04 -0.06 -0.07 -0.05 -0.11 -2.16 -1.77
(-1.32) (-0.98) (-1.42) (-0.51) (-0.23) (-0.94) (-1.78) (-2.10)
Fund only –
Benchmark only (F-B)
1.05 1.19 -1.09 0.31 -1.12 0.14 0.39 -0.79
(0.57) (0.93) (-1.10) (0.32) (-0.81) (0.08) (0.16) (-0.43)

40

Table 4. Average yearly gross returns (%) of stocks, based on the fund’s treatment of them, grouped by the fund’s benchmark
For each fund I calculate returns of 4 value weighted portfolios related to screening and weighting. Screen-out contains stocks that were screened out (same stocks
as in “Benchmark only” in Tables 2 and 3), screen-in contains stocks that were screened in (same stocks as in “Fund only” in Tables 2 and 3), overweight contains
stocks that were overweighted, and underweight contains stocks that were underweighted as compared to value weights. The overweight and underweight portfolios
contain stocks from within the entire fund, and not only those that were screened in. Then I calculate averages for each of the 4 portfolios across all the funds in a
specific style category each month. Finally, I calculate the time-series average for each measure for each fund style, and report the results. Out-In is the return of a
long-short portfolio buying stocks that were screened-out and selling the stocks that were screened-in. Over-Under is the return of a long-short portfolio buying
stocks that were overweighted and selling the stocks that were underweighted, as compared to value weights. T-statistics are calculated using standard errors of
time series mean and are presented in brackets.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 16176 2402 2869 2229 1395 932 907 993 1146

Screen-Out
12.95 12.09 14.67 13.19 12.77 11.93 9.67 10.00 10.32
(5.18) (4.84) (4.88) (4.50) (3.92) (3.00) (2.05) (2.23) (2.32)
Screen-In
12.35 12.21 12.65 15.11 11.95 11.41 7.99 9.76 7.75
(4.52) (4.40) (3.89) (5.02) (3.37) (2.84) (1.50) (2.12) (1.48)
Out-In
0.60 -0.12 2.03 -1.91 0.82 0.52 1.68 0.24 2.57
(0.92) (-0.09) (1.38) (-1.78) (0.73) (0.41) (0.77) (0.15) (1.51)
Overweight
13.28 12.98 14.62 14.44 13.36 12.21 9.44 10.12 9.38
(4.82) (4.93) (4.22) (4.72) (3.69) (3.03) (1.93) (2.12) (1.90)
Underweight
12.11 11.45 13.74 12.79 12.55 11.60 10.02 8.71 9.40
(4.64) (4.47) (4.00) (4.43) (3.53) (2.91) (2.35) (2.02) (2.14)
Over-Under
1.18 1.53 0.88 1.65 0.81 0.61 -0.58 1.41 -0.01
(1.27) (1.40) (0.63) (1.25) (0.95) (0.68) (-0.35) (0.99) (-0.01)

41

Table 4 (continued). Average yearly gross returns (%) of stocks, based on the fund’s treatment of them, grouped by the fund’s
benchmark
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 187 264 316 479 199 305 315 1238

Screen-Out
11.98 11.46 11.39 9.86 10.52 10.35 10.11 13.04
(2.08) (1.95) (1.89) (1.83) (1.94) (1.91) (1.83) (3.20)
Screen-In
12.19 12.24 10.40 9.25 7.87 10.27 10.18 12.38
(2.24) (2.15) (1.74) (1.73) (1.39) (1.84) (1.74) (2.40)
Out-In
-0.20 -0.78 0.99 0.60 2.64 0.08 -0.07 0.66
(-0.09) (-0.41) (0.62) (0.47) (1.66) (0.04) (-0.03) (0.31)
Overweight
13.76 12.57 11.51 10.51 11.56 10.74 12.37 15.41
(2.47) (2.14) (1.95) (1.91) (2.06) (1.89) (2.47) (3.38)
Underweight
12.26 11.50 10.24 8.67 8.92 9.81 10.76 12.23
(2.30) (2.08) (1.77) (1.65) (1.69) (1.80) (2.36) (3.13)
Over-Under
1.49 1.06 1.26 1.84 2.64 0.93 1.61 3.17
(1.74) (0.86) (1.19) (2.09) (1.85) (1.13) (1.31) (1.99)

42

Table 5. Average yearly excess (DGTW) returns (%) of stocks, based on the fund’s treatment of them, grouped by the fund’s
benchmark
For each fund I calculate returns of 4 value weighted portfolios related to screening and weighting, adjusting the vector of stock returns following Daniel et al.
(1997). Screen-out contains stocks that were screened out (same stocks as in “Benchmark only” in Tables 2 and 3), screen-in contains stocks that were screened in
(same stocks as in “Fund only” in Tables 2 and 3), overweight contains stocks that were overweighted, and underweight contains stocks that were underweighted
as compared to value weights. The overweight and underweight portfolios contain stocks from within the entire fund, and not only those that were screened in.
Then I calculate averages for each of the 4 portfolios across all the funds in a specific style category each month. Finally, I calculate the time-series average for
each measure for each fund style, and report the results. Out-In is the return of a long-short portfolio buying stocks that were screened-out and selling the stocks
that were screened-in. Over-Under is the return of a long-short portfolio buying stocks that were overweighted and selling the stocks that were underweighted, as
compared to value weights. T-statistics are calculated using standard errors of time series mean and are presented in brackets.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 16176 2402 2869 2229 1395 932 907 993 1146

Screen-Out
0.53 0.13 1.24 -0.29 0.83 0.30 0.35 0.66 1.01
(2.81) (0.54) (2.05) (-0.53) (1.23) (0.38) (0.64) (1.18) (1.64)
Screen-In
0.21 0.83 -0.57 1.23 0.75 0.46 -1.01 1.01 -0.91
(0.56) (1.05) (-0.87) (2.12) (1.14) (0.77) (-0.62) (1.44) (-1.00)
Out-In
0.33 -0.69 1.82 -1.51 0.08 -0.16 1.36 -0.35 1.92
(0.75) (-0.80) (1.83) (-1.85) (0.08) (-0.17) (0.75) (-0.32) (1.57)
Overweight
0.70 0.70 0.99 0.59 1.45 0.50 0.25 0.56 0.41
(1.91) (1.63) (1.55) (0.80) (2.50) (0.74) (0.38) (0.84) (0.48)
Underweight
0.30 0.05 0.46 0.17 1.12 0.62 1.51 0.27 0.79
(0.84) (0.11) (0.64) (0.26) (1.60) (1.09) (1.98) (0.33) (0.85)
Over-Under
0.40 0.65 0.54 0.42 0.33 -0.12 -1.25 0.29 -0.38
(0.89) (1.07) (0.65) (0.59) (0.66) (-0.20) (-1.21) (0.34) (-0.35)

43

Table 5 (continued). Average yearly excess (DGTW) returns (%) of stocks, based on the fund’s treatment of them, grouped by
the fund’s benchmark
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 187 264 316 479 199 305 315 1238

Screen-Out
-0.40 0.42 1.28 0.08 0.22 0.39 -0.95 -0.06
(-0.33) (0.39) (1.36) (0.09) (0.30) (0.52) (-1.00) (-0.14)
Screen-In
0.65 1.61 0.18 0.40 -0.89 0.53 -0.56 -0.86
(0.53) (2.22) (0.20) (0.40) (-0.66) (0.27) (-0.25) (-0.50)
Out-In
-1.05 -1.19 1.09 -0.31 1.12 -0.14 -0.39 0.79
(-0.57) (-0.93) (1.10) (-0.32) (0.81) (-0.08) (-0.16) (0.43)
Overweight
0.90 1.37 1.40 1.01 1.54 0.46 1.45 1.94
(1.14) (1.41) (1.32) (1.14) (1.44) (0.27) (1.42) (1.81)
Underweight
0.47 0.83 0.34 -0.03 -0.42 0.35 0.95 0.12
(0.65) (1.09) (0.31) (-0.04) (-0.41) (0.19) (0.95) (0.16)
Over-Under
0.43 0.54 1.06 1.04 1.97 0.12 0.50 1.82
(0.65) (0.71) (1.54) (2.15) (1.62) (0.19) (0.66) (1.72)

44

Table 6. Average yearly returns (%) of funds grouped by their benchmark
I first compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then calculate value weighted average across funds at
time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report
𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the sample standard deviation of 𝑀 𝑐 , assuming no serial
correlation in the returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 16176 2402 2869 2229 1395 932 907 993 1146

Screen-In
-0.19 -0.47 0.63 -0.80 1.48 -0.07 1.60 -0.76 -0.05
(-0.40) (-0.86) (0.68) (-0.88) (1.69) (-0.13) (1.79) (-0.83) (-0.05)
Screen-Out
-0.21 0.10 -1.72 1.60 -1.46 -0.18 -2.42 0.24 -0.76
(-0.61) (0.27) (-1.79) (2.12) (-2.48) (-0.25) (-2.69) (0.29) (-0.95)
Screening (screen-in +
screen-out)
-0.40 -0.37 -1.09 0.81 0.03 -0.25 -0.82 -0.52 -0.81
(-0.93) (-0.72) (-1.17) (1.14) (0.03) (-0.30) (-0.93) (-0.61) (-1.05)
Overweight
5.16 4.76 5.91 6.14 4.32 3.80 3.97 3.96 3.74
(4.84) (4.98) (4.30) (4.79) (3.58) (3.09) (2.07) (2.08) (1.86)
Underweight
4.44 3.82 4.93 5.52 4.16 3.39 3.22 3.27 3.56
(4.60) (4.18) (3.96) (4.68) (3.59) (2.81) (2.04) (2.01) (2.10)
Weighting (overweight
- underweight)
0.72 0.94 0.98 0.62 0.16 0.40 0.76 0.69 0.18
(1.83) (2.03) (1.70) (1.09) (0.38) (1.08) (1.06) (1.04) (0.21)
Active (screening +
weighting)
0.32 0.57 -0.10 1.43 0.19 0.15 -0.06 0.17 -0.63
(0.54) (0.94) (-0.09) (1.50) (0.20) (0.21) (-0.06) (0.18) (-0.69)
Benchmark
12.88 12.28 14.68 13.06 13.01 12.13 9.84 9.70 10.17
(5.17) (5.03) (4.71) (4.51) (3.96) (3.07) (2.22) (2.22) (2.34)
Total Returns (active +
benchmark)
13.20 12.85 14.58 14.49 13.19 12.27 9.78 9.86 9.54
(4.83) (4.92) (4.25) (4.84) (3.64) (3.06) (2.03) (2.11) (1.97)
Value Weighted Fund
Returns
12.48 11.91 13.59 13.86 13.03 11.87 9.03 9.17 9.36
(4.75) (4.64) (4.10) (4.80) (3.65) (2.99) (2.01) (2.08) (2.07)

45

Table 6 (continued). Average yearly gross returns (%) of funds grouped by their benchmark
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 187 264 316 479 199 305 315 1238

Screen-In
-0.34 -0.29 -0.70 -0.15 -1.00 -0.32 0.36 0.37
(-0.56) (-0.42) (-0.76) (-0.19) (-0.90) (-0.20) (0.38) (0.21)
Screen-Out
0.45 0.67 0.15 -0.25 0.37 0.25 0.20 0.73
(0.52) (0.71) (0.20) (-0.30) (0.41) (0.45) (0.18) (0.40)
Screening (screen-in +
screen-out)
0.11 0.37 -0.55 -0.40 -0.63 -0.07 0.56 1.09
(0.16) (0.40) (-0.60) (-0.50) (-0.71) (-0.04) (0.54) (0.89)
Overweight
3.66 3.48 3.16 3.33 4.27 3.38 3.60 5.20
(2.50) (1.94) (1.89) (1.94) (2.21) (2.07) (2.18) (2.97)
Underweight
3.31 3.14 2.86 2.56 3.11 2.89 3.04 3.74
(2.47) (1.95) (1.80) (1.61) (1.86) (1.89) (2.28) (2.43)
Weighting (overweight
- underweight)
0.35 0.34 0.30 0.77 1.16 0.49 0.57 1.46
(1.16) (0.76) (0.78) (1.96) (1.63) (1.58) (0.95) (2.40)
Active (screening +
weighting)
0.46 0.72 -0.25 0.37 0.53 0.42 1.13 2.56
(0.66) (0.89) (-0.33) (0.55) (0.55) (0.28) (1.03) (1.87)
Benchmark
13.01 11.50 11.32 9.71 10.64 10.32 10.41 12.70
(2.34) (1.97) (1.89) (1.80) (1.99) (1.91) (2.27) (3.30)
Total Returns (active +
benchmark)
13.46 12.21 11.07 10.08 11.17 10.74 11.54 15.26
(2.41) (2.10) (1.88) (1.84) (2.00) (1.90) (2.38) (3.44)
Value Weighted Fund
Returns
13.11 11.87 10.77 9.31 10.01 10.25 10.97 13.79
(2.40) (2.11) (1.85) (1.74) (1.88) (1.85) (2.41) (3.26)

46

Table 7. Average yearly excess (DGTW) returns (%) of funds grouped by their benchmark
I first compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above, adjusting the vector of stock returns following Daniel et
al. (1997). I then calculate average across funds at time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute time-series average for each measure
and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report 𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the
sample standard deviation of 𝑀 𝑐 , assuming no serial correlation in returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P MidCap
400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 16176 2402 2869 2229 1395 932 907 993 1146

Screen-In
0.02 -0.16 0.45 -0.10 1.22 0.01 1.40 -0.43 0.25
(0.07) (-0.47) (0.76) (-0.16) (1.96) (0.01) (2.40) (-0.57) (0.40)
Screen-Out
-0.14 0.10 -1.37 1.07 -0.88 0.06 -1.70 0.27 -0.66
(-0.61) (0.38) (-2.06) (2.05) (-1.56) (0.09) (-2.66) (0.35) (-1.16)
Screening (screen-in +
screen-out)
-0.12 -0.06 -0.92 0.97 0.34 0.06 -0.30 -0.16 -0.40
(-0.44) (-0.19) (-1.56) (1.76) (0.50) (0.10) (-0.43) (-0.27) (-0.69)
Overweight
0.35 0.30 0.57 0.36 0.38 0.16 0.35 0.19 0.19
(2.33) (1.66) (2.29) (1.16) (1.86) (0.63) (1.25) (0.82) (0.54)
Underweight
0.05 -0.13 -0.14 0.30 0.43 0.20 0.19 0.09 0.23
(0.39) (-0.82) (-0.66) (1.29) (1.70) (0.97) (0.73) (0.33) (0.68)
Weighting (overweight
- underweight)
0.30 0.44 0.71 0.06 -0.05 -0.03 0.17 0.10 -0.03
(1.69) (1.77) (2.45) (0.21) (-0.20) (-0.12) (0.42) (0.29) (-0.07)
Active (screening +
weighting)
0.18 0.38 -0.21 1.03 0.29 0.03 -0.13 -0.06 -0.43
(0.61) (1.14) (-0.35) (1.75) (0.46) (0.05) (-0.19) (-0.10) (-0.76)
Benchmark
0.53 0.30 1.19 -0.28 1.03 0.50 0.76 0.52 1.03
(2.85) (2.05) (1.99) (-0.52) (1.55) (0.64) (1.55) (0.84) (1.54)
Total Returns (active +
benchmark)
0.70 0.68 0.98 0.76 1.32 0.53 0.63 0.46 0.60
(2.32) (1.97) (1.87) (1.16) (2.28) (0.84) (1.15) (0.78) (0.83)
Value Weighted Fund
Returns
0.40 0.24 0.27 0.69 1.36 0.57 0.46 0.36 0.63
(1.48) (0.79) (0.55) (1.25) (2.21) (1.01) (0.84) (0.57) (0.94)

47

Table 7 (continued). Average yearly excess (DGTW) returns (%) of funds grouped by their benchmark
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 187 264 316 479 199 305 315 1238

Screen-In
-0.43 -0.46 -0.40 0.33 -0.81 -0.20 0.23 0.00
(-0.69) (-0.65) (-0.48) (0.47) (-0.76) (-0.13) (0.28) (-0.00)
Screen-Out
1.02 0.96 -0.05 -0.07 0.91 0.26 0.59 1.01
(1.37) (1.21) (-0.08) (-0.09) (0.93) (0.45) (0.54) (0.56)
Screening (screen-in +
screen-out)
0.59 0.50 -0.45 0.25 0.10 0.06 0.82 1.01
(0.96) (0.73) (-0.68) (0.45) (0.14) (0.04) (0.86) (1.00)
Overweight
0.14 0.33 0.40 0.37 0.57 0.17 0.42 0.64
(0.60) (1.13) (1.53) (1.59) (1.18) (0.41) (1.24) (2.13)
Underweight
0.18 0.11 0.16 -0.04 0.03 0.04 0.16 -0.07
(0.86) (0.40) (0.44) (-0.11) (0.07) (0.10) (0.58) (-0.24)
Weighting (overweight
- underweight)
-0.04 0.22 0.24 0.40 0.54 0.13 0.26 0.70
(-0.15) (0.79) (0.73) (1.94) (0.87) (0.60) (0.77) (1.96)
Active (screening +
weighting)
0.55 0.72 -0.21 0.66 0.64 0.19 1.07 1.72
(0.88) (1.07) (-0.35) (1.16) (0.76) (0.13) (1.14) (1.66)
Benchmark
0.07 0.43 1.24 0.06 0.35 0.37 -0.01 0.11
(0.07) (0.39) (1.27) (0.06) (0.50) (0.49) (-0.02) (0.26)
Total Returns (active +
benchmark)
0.62 1.15 1.03 0.72 0.99 0.55 1.06 1.82
(0.83) (1.32) (1.03) (0.85) (1.03) (0.32) (1.08) (1.99)
Value Weighted Fund
Returns
0.66 0.93 0.80 0.32 0.45 0.43 0.80 1.12
(0.91) (1.12) (0.71) (0.34) (0.52) (0.24) (0.85) (1.27)

48

Table 8. Average yearly returns (%) of funds grouped by their benchmark
I first compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then calculate value weighted average across funds at
time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report
𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the sample standard deviation of 𝑀 𝑐 , assuming no serial
correlation in the returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 16176 2402 2869 2229 1395 932 907 993 1146

Screen-In
-0.19 -0.47 0.63 -0.80 1.48 -0.07 1.60 -0.76 -0.05
(-0.40) (-0.86) (0.68) (-0.88) (1.69) (-0.13) (1.79) (-0.83) (-0.05)
Screen-Out
-0.21 0.10 -1.72 1.60 -1.46 -0.18 -2.42 0.24 -0.76
(-0.61) (0.27) (-1.79) (2.12) (-2.48) (-0.25) (-2.69) (0.29) (-0.95)
Screening (screen-in +
screen-out)
-0.40 -0.37 -1.09 0.81 0.03 -0.25 -0.82 -0.52 -0.81
(-0.93) (-0.72) (-1.17) (1.14) (0.03) (-0.30) (-0.93) (-0.61) (-1.05)
Overweight
2.06 2.10 2.36 2.41 1.81 1.49 2.22 1.55 1.31
(4.88) (5.06) (4.38) (4.76) (3.52) (3.06) (2.30) (2.08) (1.81)
Evenweight
1.97 1.69 2.41 2.59 1.10 0.84 0.74 1.52 1.66
(4.64) (4.74) (4.04) (4.58) (3.05) (2.63) (1.18) (1.90) (1.79)
Underweight
3.31 2.85 3.79 4.38 2.75 1.92 2.21 2.37 2.80
(4.48) (3.97) (3.89) (4.55) (3.47) (2.51) (1.85) (1.90) (2.08)
Weighting (over + even
- under)
0.72 0.94 0.98 0.62 0.16 0.40 0.76 0.69 0.18
(1.83) (2.03) (1.70) (1.09) (0.38) (1.08) (1.06) (1.04) (0.21)
Active (screening +
weighting)
0.32 0.57 -0.10 1.43 0.19 0.15 -0.06 0.17 -0.63
(0.54) (0.94) (-0.09) (1.50) (0.20) (0.21) (-0.06) (0.18) (-0.69)
Benchmark
12.88 12.28 14.68 13.06 13.01 12.13 9.84 9.70 10.17
(5.17) (5.03) (4.71) (4.51) (3.96) (3.07) (2.22) (2.22) (2.34)
Total Returns (active +
benchmark)
13.20 12.85 14.58 14.49 13.19 12.27 9.78 9.86 9.54
(4.83) (4.92) (4.25) (4.84) (3.64) (3.06) (2.03) (2.11) (1.97)
Value Weighted Fund
Returns
12.48 11.91 13.59 13.86 13.03 11.87 9.03 9.17 9.36
(4.75) (4.64) (4.10) (4.80) (3.65) (2.99) (2.01) (2.08) (2.07)

49

Table 8 (continued). Average yearly returns (%) of funds grouped by their benchmark
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 187 264 316 479 199 305 315 1238

Screen-In
-0.34 -0.29 -0.70 -0.15 -1.00 -0.32 0.36 0.37
(-0.56) (-0.42) (-0.76) (-0.19) (-0.90) (-0.20) (0.38) (0.21)
Screen-Out
0.45 0.67 0.15 -0.25 0.37 0.25 0.20 0.73
(0.52) (0.71) (0.20) (-0.30) (0.41) (0.45) (0.18) (0.40)
Screening (screen-in +
screen-out)
0.11 0.37 -0.55 -0.40 -0.63 -0.07 0.56 1.09
(0.16) (0.40) (-0.60) (-0.50) (-0.71) (-0.04) (0.54) (0.89)
Overweight
1.88 1.15 1.08 0.99 1.95 1.09 1.29 2.32
(2.43) (1.56) (1.88) (1.85) (2.10) (2.22) (1.65) (2.50)
Evenweight
0.27 0.74 0.47 1.20 0.82 1.02 1.51 1.81
(1.30) (2.26) (1.54) (2.13) (1.79) (2.02) (2.41) (2.96)
Underweight
1.80 1.55 1.25 1.42 1.61 1.62 2.23 2.66
(2.49) (1.80) (1.68) (1.48) (1.70) (1.86) (2.17) (2.08)
Weighting (over + even
- under)
0.35 0.34 0.30 0.77 1.16 0.49 0.57 1.46
(1.16) (0.76) (0.78) (1.96) (1.63) (1.58) (0.95) (2.40)
Active (screening +
weighting)
0.46 0.72 -0.25 0.37 0.53 0.42 1.13 2.56
(0.66) (0.89) (-0.33) (0.55) (0.55) (0.28) (1.03) (1.87)
Benchmark
13.01 11.50 11.32 9.71 10.64 10.32 10.41 12.70
(2.34) (1.97) (1.89) (1.80) (1.99) (1.91) (2.27) (3.30)
Total Returns (active +
benchmark)
13.46 12.21 11.07 10.08 11.17 10.74 11.54 15.26
(2.41) (2.10) (1.88) (1.84) (2.00) (1.90) (2.38) (3.44)
Value Weighted Fund
Returns
13.11 11.87 10.77 9.31 10.01 10.25 10.97 13.79
(2.40) (2.11) (1.85) (1.74) (1.88) (1.85) (2.41) (3.26)

50

Table 9. DGTW adjusted average yearly returns (%) of funds grouped by their benchmark
I first compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then calculate value weighted average across funds at
time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report
𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the sample standard deviation of 𝑀 𝑐 , assuming no serial
correlation in the returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 16176 2402 2869 2229 1395 932 907 993 1146

Screen-In
0.02 -0.16 0.45 -0.10 1.22 0.01 1.40 -0.43 0.25
(0.07) (-0.47) (0.76) (-0.16) (1.96) (0.01) (2.40) (-0.57) (0.40)
Screen-Out
-0.14 0.10 -1.37 1.07 -0.88 0.06 -1.70 0.27 -0.66
(-0.61) (0.38) (-2.06) (2.05) (-1.56) (0.09) (-2.66) (0.35) (-1.16)
Screening (screen-in +
screen-out)
-0.12 -0.06 -0.92 0.97 0.34 0.06 -0.30 -0.16 -0.40
(-0.44) (-0.19) (-1.56) (1.76) (0.50) (0.10) (-0.43) (-0.27) (-0.69)
Overweight
0.16 0.11 0.33 0.12 0.25 0.05 0.17 0.01 0.00
(2.27) (1.03) (2.63) (0.86) (1.96) (0.47) (0.88) (0.10) (0.00)
Evenweight
0.13 0.18 0.20 0.15 -0.06 -0.01 0.03 0.09 0.13
(1.56) (1.88) (1.47) (0.94) (-0.58) (-0.07) (0.17) (0.59) (0.66)
Underweight
-0.01 -0.15 -0.18 0.21 0.23 0.08 0.04 -0.00 0.16
(0.06) (-0.97) (-0.92) (0.98) (1.16) (0.44) (0.16) (-0.01) (0.51)
Weighting (over + even
- under)
0.30 0.44 0.71 0.06 -0.05 -0.03 0.17 0.10 -0.03
(1.69) (1.77) (2.45) (0.21) (-0.20) (-0.12) (0.42) (0.29) (-0.07)
Active (screening +
weighting)
0.18 0.38 -0.21 1.03 0.29 0.03 -0.13 -0.06 -0.43
(0.61) (1.14) (-0.35) (1.75) (0.46) (0.05) (-0.19) (-0.10) (-0.76)
Benchmark
0.53 0.30 1.19 -0.28 1.03 0.50 0.76 0.52 1.03
(2.85) (2.05) (1.99) (-0.52) (1.55) (0.64) (1.55) (0.84) (1.54)
Total Returns (active +
benchmark)
0.70 0.68 0.98 0.76 1.32 0.53 0.63 0.46 0.60
(2.32) (1.97) (1.87) (1.16) (2.28) (0.84) (1.15) (0.78) (0.83)
Value Weighted Fund
Returns
0.40 0.24 0.27 0.69 1.36 0.57 0.46 0.36 0.63
(1.48) (0.79) (0.55) (1.25) (2.21) (1.01) (0.84) (0.57) (0.94)

51

Table 9 (continued). DGTW adjusted average yearly returns (%) of funds grouped by their benchmark
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 187 264 316 479 199 305 315 1238

Screen-In
-0.43 -0.46 -0.40 0.33 -0.81 -0.20 0.23 0.00
(-0.69) (-0.65) (-0.48) (0.47) (-0.76) (-0.13) (0.28) (-0.00)
Screen-Out
1.02 0.96 -0.05 -0.07 0.91 0.26 0.59 1.01
(1.37) (1.21) (-0.08) (-0.09) (0.93) (0.45) (0.54) (0.56)
Screening (screen-in +
screen-out)
0.59 0.50 -0.45 0.25 0.10 0.06 0.82 1.01
(0.96) (0.73) (-0.68) (0.45) (0.14) (0.04) (0.86) (1.00)
Overweight
0.03 -0.06 0.11 0.04 0.42 -0.04 -0.01 0.23
(0.16) (-0.38) (0.91) (0.43) (1.18) (-0.33) (-0.04) (1.28)
Evenweight
0.01 0.17 0.15 0.30 0.19 0.13 0.31 0.34
(0.10) (1.28) (1.06) (2.56) (0.74) (0.85) (1.93) (2.11)
Underweight
0.08 -0.12 0.02 -0.07 0.07 -0.03 0.04 -0.13
(0.49) (-0.54) (0.09) (-0.32) (0.22) (-0.11) (0.19) (-0.54)
Weighting (over + even
- under)
-0.04 0.22 0.24 0.40 0.54 0.13 0.26 0.70
(-0.15) (0.79) (0.73) (1.94) (0.87) (0.60) (0.77) (1.96)
Active (screening +
weighting)
0.55 0.72 -0.21 0.66 0.64 0.19 1.07 1.72
(0.88) (1.07) (-0.35) (1.16) (0.76) (0.13) (1.14) (1.66)
Benchmark
0.07 0.43 1.24 0.06 0.35 0.37 -0.01 0.11
(0.07) (0.39) (1.27) (0.06) (0.50) (0.49) (-0.02) (0.26)
Total Returns (active +
benchmark)
0.62 1.15 1.03 0.72 0.99 0.55 1.06 1.82
(0.83) (1.32) (1.03) (0.85) (1.03) (0.32) (1.08) (1.99)
Value Weighted Fund
Returns
0.66 0.93 0.80 0.32 0.45 0.43 0.80 1.12
(0.91) (1.12) (0.71) (0.34) (0.52) (0.24) (0.85) (1.27)

52

Table 10. Quarterly transition matrix (%) of funds grouped by their benchmark (from t to t+1)
The table shows the probability of assigning a particular benchmark at t+1, given a benchmark at t.

t+1

S&P 500
S&P 500
Growth
S&P 500
Value
S&P MidCap
400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
t
S&P 500 75.2 8.6 5.6 0.2 0.1 1.6 3.2 3.3
S&P 500 Growth 5.7 87.6 2.9 1.3 0.3 0.3 0.1 0.8
S&P 500 Value 6.8 5.5 84.2 1.6 0.5 0.0 0.5 0.0
S&P MidCap 400 0.4 3.8 2.8 82.6 6.4 0.0 0.1 0.2
S&P SmallCap 600 0.1 1.0 0.9 6.8 87.6 - 0.0 0.1
Russell 1000 16.8 4.5 0.5 0.1 - 61.3 3.7 6.3
Russell 1000 Value 8.7 0.4 2.8 0.1 0.0 1.9 80.1 3.2
Russell 1000 Growth 6.0 3.4 0.1 0.2 0.0 2.1 2.6 82.2
Russell 2000 - 0.4 - 3.3 4.0 0.2 - -
Russell 2000 Value 0.1 - - 3.0 8.8 - 0.6 0.1
Russell 2000 Growth 0.0 - 0.0 3.1 6.8 - 0.3 0.2
Russell MidCap
Growth
0.3 0.7 0.5 6.3 0.3 0.3 1.5 3.7
Russell MidCap 2.0 - 0.4 4.5 0.4 - 1.2 2.4
Russell MidCap
Value
0.3 0.0 1.2 2.8 0.5 0.2 2.9 1.0
Russell 3000 3.5 15.3 2.0 2.6 4.9 4.9 0.2 1.4
Wilshire 5000 22.5 6.3 2.0 1.1 0.8 3.7 4.9 5.5

53

Table 10 (continued). Quarterly transition matrix (%) of funds grouped by their benchmark (from t to t+1)

t+1

Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell MidCap
Value
Russell
3000
Wilshire
5000
t
S&P 500 - - - 0.1 0.0 0.0 0.1 2.1
S&P 500 Growth - - - 0.1 - 0.0 0.2 0.8
S&P 500 Value - 0.0 - 0.1 - 0.0 0.0 0.7
S&P MidCap 400 0.1 0.3 0.8 1.5 0.1 0.5 0.0 0.3
S&P SmallCap 600 0.2 1.2 1.7 0.1 - 0.1 0.0 0.2
Russell 1000 - - 0.0 0.5 0.1 0.1 0.4 5.6
Russell 1000 Value 0.0 0.1 0.1 0.5 0.0 0.9 0.0 1.1
Russell 1000 Growth 0.0 0.0 0.1 1.1 0.0 0.2 0.0 2.0
Russell 2000 73.9 4.8 7.7 0.4 0.7 0.4 4.2 -
Russell 2000 Value 1.1 80.9 2.9 0.5 - 1.5 0.3 0.3
Russell 2000 Growth 1.2 1.1 85.7 1.1 - 0.2 0.1 0.1
Russell MidCap
Growth
0.0 0.1 0.9 80.7 1.2 2.2 - 1.4
Russell MidCap 1.2 - - 20.7 50.4 16.3 - 0.4
Russell MidCap
Value
0.2 0.9 0.2 2.3 1.1 84.8 0.8 0.7
Russell 3000 - - - - - - 52.3 13.0
Wilshire 5000 0.0 0.1 0.2 1.8 0.1 0.9 1.0 49.3

54

Table 11. Average yearly returns (%) of funds grouped by their benchmark, obtained using Frank Russell index construction
methodology
I first compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then calculate value weighted average across funds at
time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report
𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the sample standard deviation of 𝑀 𝑐 , assuming no serial
correlation in the returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 17115 2110 2673 1863 1122 793 307 1443 1578

Screen-In
0.11 -0.43 0.49 -1.05 1.74 -0.48 0.24 0.14 -0.14
(0.24) (-0.76) (0.53) (-1.04) (1.58) (-0.98) (0.26) (0.17) (-0.17)
Screen-Out
-0.23 0.00 -1.50 1.39 -1.50 0.01 -0.17 -0.19 0.22
(-0.86) (0.01) (-1.58) (1.74) (-2.46) (0.02) (-0.31) (-0.48) (0.53)
Screening (screen-in +
screen-out)
-0.12 -0.43 -1.02 0.34 0.24 -0.47 0.07 -0.05 0.08
(-0.33) (-0.95) (-1.19) (0.49) (0.23) (-0.63) (0.09) (-0.07) (0.11)
Overweight
5.18 4.78 6.05 6.13 4.56 3.75 7.95 5.29 5.98
(4.87) (5.01) (4.43) (4.73) (3.62) (2.99) (4.06) (4.55) (4.24)
Underweight
4.47 3.85 5.07 5.38 4.17 3.41 5.40 4.49 4.83
(4.61) (4.28) (4.11) (4.47) (3.43) (2.79) (3.20) (4.04) (3.81)
Weighting (overweight
- underweight)
0.71 0.93 0.99 0.75 0.39 0.34 2.56 0.81 1.15
(1.80) (1.87) (1.63) (1.15) (0.85) (0.89) (1.85) (1.42) (1.50)
Active (screening +
weighting)
0.59 0.50 -0.03 1.09 0.63 -0.13 2.63 0.75 1.23
(1.19) (1.14) (-0.03) (1.31) (0.57) (-0.21) (1.80) (0.87) (1.36)
Benchmark
12.62 12.28 14.68 13.06 13.01 12.13 10.71 12.16 11.30
(5.02) (5.03) (4.71) (4.51) (3.96) (3.07) (3.77) (4.84) (4.44)
Total Returns (active +
benchmark)
13.21 12.78 14.65 14.14 13.64 11.99 13.33 12.91 12.52
(4.83) (5.05) (4.36) (4.82) (3.64) (2.98) (3.99) (4.68) (4.21)
Value Weighted Fund
Returns
12.50 11.85 13.66 13.40 13.25 11.65 10.78 12.11 11.37
(4.74) (4.78) (4.22) (4.71) (3.59) (2.93) (3.60) (4.52) (4.06)

55

Table 11 (continued). Average yearly returns (%) of funds grouped by their benchmark, obtained using Frank Russell index
construction methodology
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 243 598 790 1048 323 913 42 1269

Screen-In
1.63 1.51 0.17 -0.37 -2.44 1.31 1.63 -0.76
(1.28) (2.27) (0.17) (-0.41) (-1.39) (1.45) (0.83) (-1.27)
Screen-Out
1.67 -1.43 -0.24 -0.08 0.84 -0.92 -0.79 0.68
(1.70) (-2.79) (-0.50) (-0.18) (0.75) (-1.89) (-0.81) (1.51)
Screening (screen-in +
screen-out)
3.30 0.08 -0.07 -0.45 -1.61 0.39 0.84 -0.08
(2.79) (0.13) (-0.08) (-0.53) (-1.11) (0.51) (0.48) (-0.15)
Overweight
3.52 4.59 3.97 4.14 5.03 5.39 9.34 6.01
(3.45) (4.89) (3.67) (3.52) (3.61) (5.37) (2.89) (3.39)
Underweight
5.17 4.42 3.66 4.03 4.18 4.49 9.99 4.24
(4.68) (4.71) (3.34) (3.56) (2.96) (4.87) (3.29) (2.90)
Weighting (overweight
- underweight)
-1.65 0.18 0.32 0.11 0.85 0.90 -0.66 1.77
(-2.50) (0.55) (1.02) (0.28) (0.91) (2.17) (-0.24) (2.43)
Active (screening +
weighting)
1.65 0.26 0.24 -0.34 -0.75 1.30 0.19 1.69
(1.47) (0.40) (0.30) (-0.39) (-0.52) (1.51) (0.08) (2.63)
Benchmark
12.20 14.39 12.45 12.22 13.53 13.44 14.88 11.91
(3.92) (4.97) (3.99) (4.30) (4.96) (5.03) (3.68) (3.07)
Total Returns (active +
benchmark)
13.85 14.65 12.70 11.88 12.78 14.73 15.06 13.60
(4.07) (5.21) (3.56) (3.55) (3.86) (5.34) (3.16) (3.25)
Value Weighted Fund
Returns
15.49 14.47 12.38 11.77 11.93 13.83 15.72 11.83
(4.46) (5.15) (3.46) (3.56) (3.55) (5.13) (3.49) (3.04)

56

Table 12. Average yearly returns (%) of funds grouped by their benchmark, including 48 industry portfolios
I first compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then calculate value weighted average across funds at
time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report
𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the sample standard deviation of 𝑀 𝑐 , assuming no serial
correlation in the returns.
Overall
Broad
index only
Industry
index only
S&P
500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell
1000
Russell
1000 Value
Number of funds 27439 9993 17446 1744 2051 1054 600 530 166 648

Screen-In
0.22 0.01 0.47 -0.46 -0.03 -1.05 1.29 -0.43 -0.50 0.16
(0.57) (0.01) (1.26) (-0.85) (-0.03) (-0.89) (1.19) (-0.84) (-0.67) (0.19)
Screen-Out
-0.28 -0.28 -0.32 0.03 -1.23 1.51 -1.70 0.01 0.03 -0.21
(-1.26) (-0.89) (-0.79) (0.13) (-1.27) (1.83) (-3.16) (0.02) (0.07) (-0.43)
Screening (screen-in +
screen-out)
-0.06 -0.27 0.15 -0.43 -1.25 0.46 -0.41 -0.41 -0.47 -0.06
(-0.18) (-0.65) (0.37) (-0.99) (-1.39) (0.57) (-0.38) (-0.54) (-0.73) (-0.08)
Overweight
5.33 5.13 5.63 4.57 5.99 6.21 3.85 3.69 7.39 5.94
(4.82) (4.97) (4.68) (4.97) (4.53) (4.81) (3.36) (3.03) (3.50) (4.39)
Underweight
4.65 4.36 5.06 3.68 5.08 5.40 3.59 3.26 5.09 4.90
(4.66) (4.58) (4.73) (4.24) (4.17) (4.51) (3.15) (2.71) (2.89) (3.78)
Weighting (overweight
- underweight)
0.68 0.77 0.58 0.89 0.91 0.81 0.26 0.43 2.30 1.04
(1.58) (1.68) (1.29) (1.85) (1.30) (1.14) (0.62) (1.22) (1.56) (1.28)
Active (screening +
weighting)
0.62 0.49 0.73 0.46 -0.34 1.27 -0.15 0.02 1.83 0.98
(1.53) (1.17) (1.45) (1.13) (-0.40) (1.60) (-0.14) (0.03) (1.29) (1.37)
Benchmark
12.59 12.62 12.80 12.28 14.68 13.06 13.01 12.13 10.35 11.07
(4.94) (5.02) (4.88) (5.03) (4.71) (4.51) (3.96) (3.07) (3.42) (4.34)
Total Returns (active +
benchmark)
13.21 13.12 13.53 12.73 14.34 14.32 12.86 12.14 12.18 12.05
(4.83) (4.94) (4.76) (5.06) (4.42) (4.90) (3.50) (2.99) (3.54) (4.46)
Value Weighted Fund
Returns
12.53 12.35 12.95 11.84 13.42 13.51 12.60 11.71 9.88 11.01
(4.78) (4.80) (4.81) (4.78) (4.27) (4.77) (3.44) (2.90) (3.22) (4.20)

57

Table 12 (continued). Average yearly returns (%) of funds grouped by their benchmark, including 48 industry portfolios

Russell 1000
Growth
Russell
2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell
3000
Wilshire
5000
Number of funds 1032 49 204 379 328 44 139 20 1005

Screen-In
0.41 0.28 2.15 0.86 -0.63 1.69 2.65 2.25 -1.57
(0.50) (0.51) (3.37) (0.95) (-0.70) (1.56) (4.23) (1.83) (-2.38)
Screen-Out
0.14 0.70 -1.05 -0.97 -0.30 -0.78 -1.84 -0.16 0.28
(0.32) (0.95) (-2.37) (-1.65) (-0.63) (-0.94) (-3.46) (-0.24) (0.58)
Screening (screen-in +
screen-out)
0.54 0.97 1.10 -0.11 -0.93 0.92 0.81 2.09 -1.29
(0.84) (1.55) (1.94) (-0.11) (-1.08) (0.79) (1.21) (1.66) (-2.26)
Overweight
6.33 3.74 4.18 3.92 4.08 4.70 5.66 10.35 5.67
(4.55) (3.73) (4.41) (3.51) (3.43) (2.73) (4.64) (3.31) (3.30)
Underweight
5.14 4.36 4.64 3.37 3.73 4.28 5.15 11.33 3.68
(4.13) (4.19) (4.92) (2.85) (3.29) (2.40) (4.81) (4.36) (2.60)
Weighting (overweight
- underweight)
1.20 -0.61 -0.46 0.56 0.35 0.42 0.51 -0.98 2.00
(1.49) (-1.06) (-1.06) (1.18) (0.65) (0.39) (0.91) (-0.41) (2.76)
Active (screening +
weighting)
1.74 0.36 0.64 0.45 -0.58 1.34 1.32 1.11 0.70
(2.24) (0.55) (1.12) (0.51) (-0.62) (1.01) (1.71) (0.47) (1.30)
Benchmark
11.30 12.69 14.14 12.55 12.03 11.67 13.01 16.80 11.91
(4.44) (3.82) (4.80) (3.99) (4.20) (3.11) (4.34) (4.06) (3.07)
Total Returns (active +
benchmark)
13.03 13.05 14.78 13.00 11.46 13.01 14.33 17.91 12.61
(4.59) (3.99) (5.04) (3.65) (3.49) (3.43) (4.63) (3.77) (3.06)
Value Weighted Fund
Returns
11.84 13.67 15.24 12.44 11.11 12.59 13.82 18.89 10.62
(4.42) (4.19) (5.21) (3.47) (3.42) (3.25) (4.69) (4.42) (2.77)

58

Table 12 (continued). Average yearly returns (%) of funds grouped by their benchmark, including 48 industry portfolios
Agric Food Soda Beer Smoke Toys Fun Books Hshld Clths Hlth MedEq
Number of funds 123 220 81 256 9 269 251 239 364 271 549 334

Screen-In
-0.41 -1.47 -1.45 2.84 -8.06 -1.31 1.42 0.14 -0.88 -5.28 2.47 -0.18
(-0.11) (-0.62) (-1.45) (0.84) (-0.42) (-0.37) (0.59) (0.07) (-0.44) (-1.81) (1.00) (-0.10)
Screen-Out
-2.51 -0.08 3.59 -2.67 6.33 0.40 -2.65 0.48 -2.86 4.95 -0.79 -1.84
(-0.60) (-0.03) (1.27) (-0.65) (0.42) (0.11) (-0.85) (0.20) (-1.32) (1.41) (-0.24) (-0.77)
Screening (screen-in +
screen-out)
-2.92 -1.55 2.14 0.17 -1.73 -0.91 -1.23 0.63 -3.74 -0.33 1.68 -2.03
(-0.90) (-0.75) (0.77) (0.06) (-0.11) (-0.31) (-0.48) (0.35) (-2.15) (-0.13) (0.65) (-1.10)
Overweight
4.78 7.32 5.96 5.99 -0.56 4.39 7.48 6.35 6.38 4.58 5.58 6.68
(3.27) (5.71) (4.42) (4.12) (-0.08) (3.20) (5.26) (5.05) (4.56) (3.84) (4.29) (4.77)
Underweight
3.98 4.99 4.96 5.68 7.84 5.55 5.15 5.80 4.11 4.89 4.38 5.78
(2.54) (4.04) (3.96) (3.41) (1.40) (3.91) (3.68) (4.55) (3.11) (4.14) (3.59) (4.63)
Weighting (overweight
- underweight)
0.80 2.33 1.01 0.31 -8.40 -1.16 2.33 0.55 2.27 -0.31 1.20 0.89
(0.67) (2.74) (0.95) (0.24) (-1.20) (-1.26) (2.35) (0.59) (2.50) (-0.42) (1.59) (1.07)
Active (screening +
weighting)
-2.11 0.78 3.15 0.48 -10.13 -2.06 1.10 1.18 -1.48 -0.64 2.88 -1.13
(-0.68) (0.37) (1.15) (0.18) (-0.67) (-0.71) (0.45) (0.67) (-0.85) (-0.25) (1.16) (-0.57)
Benchmark
14.66 15.19 11.14 13.44 17.91 14.21 15.89 13.92 15.33 13.58 12.16 16.03
(3.82) (5.86) (3.07) (4.43) (1.48) (3.18) (4.01) (4.54) (5.50) (3.72) (3.14) (5.55)
Total Returns (active +
benchmark)
12.55 15.97 14.29 13.92 7.78 12.15 16.99 15.09 13.85 12.95 15.04 14.90
(3.53) (5.80) (4.48) (4.15) (0.51) (3.41) (5.00) (5.02) (4.84) (4.01) (4.45) (4.80)
Value Weighted Fund
Returns
11.75 13.64 13.28 13.61 16.17 13.30 14.67 14.54 11.58 13.25 13.84 14.01
(3.30) (5.05) (4.33) (3.81) (1.13) (3.73) (4.43) (4.90) (4.17) (4.16) (4.22) (4.76)

59

Table 12 (continued). Average yearly returns (%) of funds grouped by their benchmark, including 48 industry portfolios
Drugs Chems Rubbr Txtls BldMt Cnstr Steel FabPr Mach ElcEq Autos Aero
Number of funds 873 318 267 150 368 412 551 123 421 425 292 285

Screen-In
-0.37 0.65 -0.05 -3.30 -1.77 1.50 -1.34 -5.26 0.23 -1.58 -0.76 0.32
(-0.54) (0.59) (-0.01) (-1.18) (-1.11) (0.61) (-0.67) (-0.95) (0.12) (-0.49) (-0.35) (0.30)
Screen-Out
-0.43 0.03 0.59 4.57 2.46 1.42 3.33 6.01 1.19 1.99 0.14 -1.42
(-0.29) (0.02) (0.15) (1.14) (1.19) (0.51) (1.19) (1.01) (0.55) (0.57) (0.06) (-0.67)
Screening (screen-in +
screen-out)
-0.81 0.68 0.54 1.27 0.69 2.92 1.99 0.75 1.42 0.41 -0.62 -1.10
(-0.60) (0.53) (0.22) (0.43) (0.42) (1.30) (0.91) (0.22) (0.93) (0.19) (-0.31) (-0.61)
Overweight
6.63 6.15 3.95 6.39 5.53 5.72 6.66 5.19 5.89 6.79 5.69 5.91
(5.05) (4.57) (3.02) (4.38) (4.19) (4.41) (4.68) (3.03) (3.88) (4.46) (3.27) (4.60)
Underweight
5.69 6.16 4.07 4.30 5.15 5.64 5.54 3.41 4.92 5.63 5.32 4.75
(5.27) (5.19) (3.17) (2.84) (4.27) (4.44) (4.23) (2.01) (3.75) (4.06) (3.02) (3.99)
Weighting (overweight
- underweight)
0.93 -0.01 -0.13 2.09 0.38 0.08 1.12 1.78 0.97 1.16 0.37 1.16
(1.24) (-0.01) (-0.17) (1.81) (0.53) (0.12) (1.40) (1.32) (1.14) (1.16) (0.37) (1.42)
Active (screening +
weighting)
0.13 0.67 0.41 3.35 1.07 3.00 3.10 2.54 2.39 1.57 -0.25 0.06
(0.08) (0.52) (0.17) (1.19) (0.71) (1.38) (1.41) (0.76) (1.64) (0.79) (-0.12) (0.03)
Benchmark
14.36 13.92 9.94 9.73 12.79 11.85 11.27 8.21 11.37 14.90 14.38 14.38
(5.30) (4.48) (2.48) (2.13) (3.81) (2.90) (2.55) (1.74) (3.08) (3.71) (3.50) (4.09)
Total Returns (active +
benchmark)
14.48 14.59 10.35 13.08 13.86 14.85 14.38 10.75 13.76 16.47 14.13 14.44
(5.42) (4.90) (3.00) (3.69) (4.54) (4.43) (4.28) (2.58) (4.23) (4.39) (3.86) (5.04)
Value Weighted Fund
Returns
13.55 14.60 10.47 10.99 13.49 14.77 13.26 8.96 12.79 15.31 13.76 13.27
(5.50) (5.17) (3.06) (3.04) (4.62) (4.47) (4.09) (2.19) (4.20) (4.28) (3.79) (4.80)

60

Table 12 (continued). Average yearly returns (%) of funds grouped by their benchmark, including 48 industry portfolios
Ships Guns Gold Mines Coal Oil Util Telcm PerSv BusSv Comps Chips
Number of funds 32 47 460 398 26 786 219 541 424 824 867 745

Screen-In
-11.95 5.98 -0.41 0.90 -4.83 -0.24 -1.01 1.52 1.39 1.85 3.93 2.01
(-1.03) (0.82) (-0.22) (0.39) (-0.43) (-0.33) (-1.45) (0.75) (0.71) (1.01) (1.87) (1.00)
Screen-Out
6.84 0.67 1.09 0.07 -4.19 0.68 1.04 -1.06 0.53 -3.27 -4.66 -2.39
(0.52) (0.06) (0.22) (0.02) (-0.23) (0.43) (0.95) (-0.58) (0.21) (-1.67) (-1.67) (-0.98)
Screening (screen-in +
screen-out)
-5.10 6.64 0.68 0.98 -9.02 0.44 0.03 0.45 1.93 -1.42 -0.73 -0.38
(-0.45) (0.54) (0.16) (0.38) (-0.74) (0.31) (0.03) (0.26) (0.82) (-1.20) (-0.36) (-0.25)
Overweight
1.27 11.06 4.42 5.77 1.87 5.51 4.76 4.81 4.39 5.76 5.75 5.74
(0.25) (1.91) (2.88) (4.20) (0.33) (4.16) (5.74) (3.66) (3.50) (3.20) (3.21) (3.12)
Underweight
1.94 11.25 4.24 4.06 2.78 4.82 4.07 4.62 4.49 5.67 4.36 5.45
(0.46) (2.22) (3.37) (3.05) (0.49) (4.53) (4.86) (3.78) (3.94) (3.48) (2.71) (3.41)
Weighting (overweight
- underweight)
-0.67 -0.18 0.17 1.71 -0.91 0.69 0.69 0.19 -0.10 0.08 1.39 0.29
(-0.22) (-0.05) (0.18) (1.90) (-0.32) (1.05) (1.68) (0.31) (-0.18) (0.11) (1.21) (0.35)
Active (screening +
weighting)
-5.77 6.46 0.85 2.68 -9.92 1.13 0.72 0.64 1.82 -1.33 0.67 -0.09
(-0.45) (0.53) (0.20) (1.10) (-0.82) (0.84) (0.72) (0.35) (0.78) (-1.02) (0.33) (-0.07)
Benchmark
7.35 22.30 9.36 9.08 10.53 11.40 11.35 11.25 9.96 15.13 12.82 13.19
(0.46) (1.33) (1.52) (2.07) (0.52) (3.54) (4.76) (3.75) (2.82) (3.94) (3.20) (3.00)
Total Returns (active +
benchmark)
1.58 28.76 10.21 11.76 0.61 12.53 12.07 11.89 11.78 13.80 13.48 13.09
(0.13) (2.33) (2.85) (3.59) (0.04) (4.24) (5.43) (3.89) (3.64) (3.52) (3.65) (3.35)
Value Weighted Fund
Returns
2.25 28.94 10.04 10.05 1.51 11.84 11.38 11.70 11.88 13.71 12.09 12.81
(0.20) (2.53) (3.00) (3.14) (0.11) (4.42) (5.08) (3.93) (3.81) (3.65) (3.42) (3.48)

61

Table 12 (continued). Average yearly returns (%) of funds grouped by their benchmark, including 48 industry portfolios
LabEq Paper Boxes Trans Whlsl Rtail Meal Banks Insur RlEst Fin Other
Number of funds 361 341 68 278 495 539 323 914 503 288 352 164

Screen-In
-3.69 -1.03 1.05 3.75 -0.48 2.40 1.10 2.45 0.37 4.32 0.17 -1.03
(-1.22) (-0.67) (0.20) (1.72) (-0.31) (2.30) (0.60) (2.00) (0.37) (1.66) (0.11) (-0.41)
Screen-Out
-1.87 2.37 -1.52 -2.23 0.36 -3.40 -1.17 -2.78 -0.11 -1.31 0.65 2.48
(-0.61) (1.14) (-0.28) (-1.01) (0.20) (-1.94) (-0.52) (-1.57) (-0.06) (-0.46) (0.67) (0.65)
Screening (screen-in +
screen-out)
-5.56 1.34 -0.47 1.53 -0.12 -1.00 -0.07 -0.33 0.26 3.01 0.82 1.44
(-2.14) (0.86) (-0.12) (0.79) (-0.09) (-0.66) (-0.04) (-0.20) (0.18) (1.39) (0.55) (0.43)
Overweight
3.74 5.29 4.41 5.04 5.06 6.45 4.88 6.14 6.22 4.79 5.40 4.38
(2.43) (4.07) (2.07) (3.92) (4.02) (4.78) (3.71) (5.05) (5.09) (4.05) (5.19) (2.96)
Underweight
2.73 4.43 4.60 4.71 4.78 5.84 5.06 5.20 4.79 5.36 5.40 3.76
(1.67) (3.81) (2.54) (3.23) (4.04) (4.84) (4.05) (4.65) (3.91) (3.87) (5.24) (2.78)
Weighting (overweight
- underweight)
1.02 0.87 -0.19 0.33 0.27 0.61 -0.18 0.94 1.43 -0.58 0.01 0.62
(0.98) (1.14) (-0.12) (0.33) (0.41) (0.91) (-0.21) (1.65) (2.09) (-0.55) (0.01) (0.62)
Active (screening +
weighting)
-4.54 2.21 -0.66 1.86 0.15 -0.39 -0.25 0.61 1.69 2.44 0.83 2.07
(-1.68) (1.43) (-0.17) (1.16) (0.11) (-0.26) (-0.15) (0.36) (1.24) (1.28) (0.55) (0.61)
Benchmark
12.74 10.61 13.38 12.42 13.02 15.46 13.53 13.67 12.73 10.43 14.27 7.82
(2.65) (3.31) (2.79) (3.83) (4.48) (5.03) (4.55) (3.84) (4.20) (3.05) (4.71) (1.77)
Total Returns (active +
benchmark)
8.20 12.82 12.72 14.27 13.17 15.07 13.28 14.28 14.42 12.86 15.10 9.88
(2.17) (4.30) (2.73) (4.37) (4.24) (5.16) (4.49) (4.85) (5.16) (4.12) (5.05) (2.95)
Value Weighted Fund
Returns
7.18 11.95 12.90 13.94 12.90 14.46 13.46 13.34 12.99 13.44 15.09 9.26
(1.87) (4.22) (2.96) (4.11) (4.29) (5.23) (4.68) (4.73) (4.70) (4.18) (5.07) (2.88)

62

Table 13. Average Active Share measure for each category of funds for three different sets of benchmarks
The table displays the average active share measure for each category of funds: first row displays average values for original benchmark assignments, the second
row displays average measure for the same categories, but with added 48 industry benchmarks, and the third row displays average active share values when using
Russell methodology to construct benchmarks. Standard deviations are displayed in the brackets.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
16 benchmarks
0.81 0.70 0.79 0.89 0.91 0.92 0.62 0.79 0.80
(0.17) (0.23) (0.14) (0.09) (0.13) (0.13) (0.23) (0.14) (0.12)
+48 industry portfolios
0.74 0.64 0.72 0.81 0.83 0.88 0.80 0.82 0.80
(0.18) (0.21) (0.12) (0.08) (0.18) (0.15) (0.13) (0.08) (0.10)
Russell methodology
0.82 0.68 0.78 0.87 0.89 0.91 0.84 0.91 0.86
(0.17) (0.21) (0.14) (0.08) (0.14) (0.13) (0.12) (0.07) (0.09)

Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
16 benchmarks
0.62 0.88 0.90 0.88 0.83 0.92 0.61 0.77
(0.29) (0.11) (0.07) (0.09) (0.15) (0.06) (0.30) (0.22)
+48 industry portfolios
0.61 0.89 0.86 0.87 0.80 0.88 0.61 0.64
(0.17) (0.11) (0.15) (0.06) (0.14) (0.04) (0.19) (0.20)
Russell methodology
0.80 0.95 0.93 0.92 0.95 0.96 0.68 0.71
(0.22) (0.08) (0.11) (0.06) (0.09) (0.04) (0.22) (0.22)

63

Table 14. Panel regression results of regressing stock returns on lagged screening weights of those stocks in funds, grouped by
their benchmark
Regressions with stock fixed effects and adjusted for Stambaugh (1999) bias following Hjalmarsson (2010), with standard errors clustered by time. I run the panel
regression for each fund separately and collect betas and their t-statistics. I report the averages of betas and the t-statstics for the point estimates in the first row. In
the subsequent three rows I report the percentage of collected t-statistics that fall into the corresponding range. For example, 91.91% of t-statistics for overall results
(1 column) fall into the [-1.96;1.96] range, meaning that only 8.09% of the beta coefficients are significant at 95% level.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Beta adjusted
0.07 -0.32 -0.07 0.05 0.16 0.24 0.61 0.96 -0.44
(1.24) (-3.30) (-1.80) (0.65) (0.69) (1.02) (1.43) (3.61) (-1.18)
% ∈ [-1.65;1.65] 87.24 88.48 89.23 88.20 78.62 80.12 91.19 91.42 90.50
% ∈ [-1.96;1.96] 91.91 92.47 93.11 93.28 85.93 87.02 94.30 94.92 94.67
% ∈ [-2.58;2.58] 96.90 97.28 97.49 97.38 94.12 94.88 98.70 98.87 98.16

Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Beta adjusted
25.31 -2.32 5.55 -0.19 -0.58 0.13 8.47 -0.16
(1.20) (-1.75) (3.12) (-0.50) (-0.20) (0.17) (1.79) (-0.44)
% ∈ [-1.65;1.65] 83.33 84.88 86.52 88.28 86.59 80.23 92.92 86.35
% ∈ [-1.96;1.96] 87.96 90.12 90.78 94.02 90.24 86.05 96.70 90.59
% ∈ [-2.58;2.58] 92.59 94.19 97.87 96.17 95.12 94.96 97.64 96.13

64

Table 15. Panel regression results of regressing DGTW adjusted stock returns on lagged screening weights of those stocks in
funds, grouped by their benchmark
Regressions with stock fixed effects and adjusted for Stambaugh (1999) bias following Hjalmarsson (2010), with standard errors clustered by time. I run the panel
regression for each fund separately and collect betas and their t-statistics. I report the averages of betas and the t-statstics for the point estimates in the first row. In
the subsequent three rows I report the percentage of collected t-statistics that fall into the corresponding range. For example, 93.00% of t-statistics for overall results
(1 column) fall into the [-1.96;1.96] range, meaning that only 7.00% of the beta coefficients are significant at 95% level.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Beta adjusted
0.01 -0.02 -0.06 -0.09 0.28 0.10 0.62 0.41 -0.64
(0.14) (-0.33) (-2.04) (-1.54) (1.65) (0.70) (1.46) (1.83) (-2.10)
% ∈ [-1.65;1.65] 87.74 88.20 88.39 88.00 80.84 82.26 89.12 88.49 90.79
% ∈ [-1.96;1.96] 93.00 93.51 93.49 93.58 87.52 89.52 94.17 94.13 95.35
% ∈ [-2.58;2.58] 97.66 98.41 97.94 97.38 95.15 95.83 98.32 98.31 98.93

Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Beta adjusted
-14.20 0.19 2.25 -0.09 -1.60 0.50 0.71 -0.04
(-1.42) (0.20) (2.08) (-0.31) (-0.77) (0.82) (0.21) (-0.20)
% ∈ [-1.65;1.65] 87.96 87.79 89.32 89.71 91.46 84.50 93.87 91.24
% ∈ [-1.96;1.96] 93.52 93.02 95.02 94.02 96.34 87.98 96.70 94.56
% ∈ [-2.58;2.58] 96.30 96.51 98.22 98.80 98.78 96.12 99.06 97.97

65

Table 16. Panel regression results of regressing stock returns on lagged screening weights of those stocks in funds, controlling
for Carhart 4 factors and illiquidity factor, grouped by their benchmark
Regressions with stock fixed effects and standard errors clustered by time. I run the panel regression for each fund separately and collect betas and their t-statistics.
I report the averages of betas and the t-statstics for the point estimates in the first row. In the subsequent three rows I report the percentage of collected t-statistics
that fall into the corresponding range. For example, 84.60% of t-statistics for overall results (1 column) fall into the [-1.96;1.96] range, meaning that only 15.40%
of the beta coefficients are significant at 95% level.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Beta adjusted
0.14 0.00 0.08 0.02 -0.03 0.24 0.86 0.15 0.45
(15.09) (-0.48) (7.82) (1.99) (-0.86) (6.14) (8.52) (5.02) (7.26)
% ∈ [-1.65;1.65] 77.79 83.07 81.46 84.78 65.73 59.43 76.32 86.02 73.01
% ∈ [-1.96;1.96] 84.60 89.77 88.28 90.64 72.67 68.41 85.18 91.31 81.09
% ∈ [-2.58;2.58] 92.73 96.72 95.72 97.12 83.57 81.27 93.49 97.86 90.93

Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Beta adjusted
0.16 0.50 0.02 0.38 0.77 0.59 1.64 0.13
(0.34) (2.52) (0.20) (4.79) (1.44) (3.68) (5.17) (2.84)
% ∈ [-1.65;1.65] 81.71 71.52 63.60 68.54 82.46 75.53 63.58 82.46
% ∈ [-1.96;1.96] 89.02 80.38 67.82 74.42 89.47 84.81 66.47 88.78
% ∈ [-2.58;2.58] 92.68 87.97 78.54 84.91 94.74 91.56 79.19 94.59

66

Table 17. Panel regression results of regressing DGTW adjusted stock returns on lagged screening weights of those stocks in
funds, controlling for Carhart 4 factors and illiquidity factor, grouped by their benchmark
Regressions with stock fixed effects and standard errors clustered by time. I run the panel regression for each fund separately and collect betas and their t-statistics.
I report the averages of betas and the t-statstics for the point estimates in the first row. In the subsequent three rows I report the percentage of collected t-statistics
that fall into the corresponding range. For example, 84.27% of t-statistics for overall results (1 column) fall into the [-1.96;1.96] range, meaning that only 15.73%
of the beta coefficients are significant at 95% level.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Beta adjusted
0.16 0.01 0.05 0.02 -0.01 0.25 1.01 0.20 0.52
(18.09) (0.76) (6.93) (2.19) (-0.30) (6.70) (9.18) (6.70) (7.80)
% ∈ [-1.65;1.65] 77.80 82.88 81.43 84.22 67.05 56.80 77.56 86.02 75.85
% ∈ [-1.96;1.96] 84.27 89.29 87.66 89.82 74.48 66.79 85.18 90.93 82.40
% ∈ [-2.58;2.58] 92.26 95.42 95.40 95.58 84.64 79.40 93.77 96.47 92.35

Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Beta adjusted
0.39 0.60 0.04 0.44 0.83 0.59 1.69 0.22
(0.85) (2.92) (0.40) (5.46) (1.56) (4.05) (5.51) (5.38)
% ∈ [-1.65;1.65] 79.27 72.78 63.22 67.52 87.72 73.84 65.90 81.86
% ∈ [-1.96;1.96] 86.59 80.38 68.58 73.66 89.47 82.28 69.94 86.67
% ∈ [-2.58;2.58] 92.68 87.97 78.54 84.65 94.74 94.94 82.66 93.49

67

Table 18. Average yearly returns (%) of funds grouped by their benchmark, obtained using Frank Russell index construction
methodology for years 1980-1998
I first compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then calculate value weighted average across funds at
time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report
𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the sample standard deviation of 𝑀 𝑐 , assuming no serial
correlation in the returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 11447 2394 1675 784 649 120 937 894 176

Screen-In
-0.05 0.94 -1.48 2.17 -0.92 2.10 -1.00 0.16 0.81
(-0.06) (0.85) (-1.23) (0.91) (-0.65) (1.16) (-0.88) (0.13) (0.38)
Screen-Out
-0.60 -3.24 2.09 -1.46 1.09 -1.50 0.18 -0.79 1.89
(-1.38) (-3.18) (2.05) (-1.03) (0.76) (-1.25) (0.31) (-1.29) (1.18)
Screening (screen-in +
screen-out)
-0.65 -2.30 0.61 0.71 0.18 0.60 -0.82 -0.63 2.70
(-1.01) (-2.22) (0.76) (0.38) (0.14) (0.41) (-0.89) (-0.57) (1.48)
Overweight
6.67 7.32 6.68 5.94 5.15 9.73 6.19 7.41 4.22
(4.11) (4.47) (4.16) (2.52) (1.64) (3.04) (3.54) (3.52) (2.60)
Underweight
6.47 6.81 6.85 7.45 6.28 9.35 6.25 7.05 6.38
(4.42) (4.61) (4.69) (3.31) (2.05) (3.58) (3.85) (3.83) (3.56)
Weighting (overweight
- underweight)
0.20 0.51 -0.17 -1.51 -1.14 0.38 -0.06 0.36 -2.16
(0.32) (0.74) (-0.23) (-1.54) (-1.11) (0.19) (-0.07) (0.32) (-1.81)
Active (screening +
weighting)
-0.45 -1.79 0.44 -0.79 -0.96 0.98 -0.89 -0.27 0.54
(-0.55) (-1.54) (0.46) (-0.40) (-0.80) (0.42) (-0.85) (-0.21) (0.30)
Benchmark
17.44 20.52 15.68 18.26 16.91 17.01 16.10 16.34 13.87
(4.96) (5.68) (4.58) (3.31) (1.92) (3.75) (4.81) (4.49) (3.21)
Total Returns (active +
benchmark)
16.99 18.73 16.12 17.47 15.95 17.99 15.21 16.08 14.41
(4.30) (4.71) (4.45) (2.73) (1.73) (3.14) (3.98) (3.72) (2.89)
Value Weighted Fund
Returns
16.79 18.22 16.29 18.97 17.09 17.62 15.28 15.72 16.57
(4.44) (4.77) (4.68) (3.02) (1.87) (3.50) (4.17) (3.92) (3.23)

68

Table 18 (continued). Average yearly returns (%) of funds grouped by their benchmark, obtained using Frank Russell index
construction methodology for years 1980-1998
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Number of funds 450 637 716 172 575 16 42

Screen-In
1.60 1.32 0.22 -6.14 -1.31 3.13 1.63
(1.74) (0.90) (0.15) (-2.03) (-0.99) (1.24) (0.83)
Screen-Out
-2.14 -0.89 0.98 1.32 -0.43 -1.68 -0.79
(-2.54) (-1.16) (1.34) (0.70) (-0.53) (-1.74) (-0.81)
Screening (screen-in +
screen-out)
-0.54 0.44 1.20 -4.82 -1.73 1.45 0.84
(-0.61) (0.33) (0.93) (-1.93) (-1.67) (0.54) (0.48)
Overweight
5.50 4.87 5.38 5.98 6.08 12.16 9.34
(4.24) (3.12) (2.97) (2.58) (3.89) (2.25) (2.89)
Underweight
5.19 4.93 6.20 5.26 5.09 15.90 9.99
(3.97) (3.10) (3.63) (2.37) (3.56) (3.66) (3.29)
Weighting (overweight
- underweight)
0.31 -0.06 -0.82 0.72 0.99 -3.74 -0.66
(0.62) (-0.14) (-1.34) (0.45) (1.32) (-0.87) (-0.24)
Active (screening +
weighting)
-0.23 0.37 0.38 -4.09 -0.75 -2.30 0.19
(-0.26) (0.31) (0.27) (-1.56) (-0.60) (-0.57) (0.08)
Benchmark
17.00 15.05 14.90 17.29 16.17 20.83 14.88
(4.42) (3.44) (3.71) (4.33) (4.49) (3.65) (3.68)
Total Returns (active +
benchmark)
16.77 15.42 15.29 13.20 15.42 18.54 15.06
(4.42) (3.02) (3.12) (2.56) (3.77) (2.45) (3.16)
Value Weighted Fund
Returns
16.46 15.49 16.10 12.48 14.43 22.28 15.06
(4.32) (3.01) (3.37) (2.42) (3.60) (3.42) (3.16)

69

Table 19. DGTW adjusted average yearly returns (%) of funds grouped by their benchmark, obtained using Frank Russell
index construction methodology for years 1980-1998
I first compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then calculate value weighted average across funds at
time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report
𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the sample standard deviation of 𝑀 𝑐 , assuming no serial
correlation in the returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 11447 2394 1675 784 649 120 937 894 176

Screen-In
0.18 -0.28 0.60 -0.44 0.94 -1.30 0.53 -0.49 0.49
(0.35) (-0.53) (0.85) (-0.60) (0.50) (-1.06) (0.46) (-0.55) (0.67)
Screen-Out
-0.45 -0.12 -2.30 1.31 -1.43 -0.46 -0.23 0.10 -0.64
(-1.77) (-0.50) (-3.18) (2.11) (-1.14) (-0.36) (-0.26) (0.24) (-1.68)
Screening (screen-in +
screen-out)
-0.27 -0.40 -1.70 0.87 -0.50 -1.75 0.30 -0.39 -0.14
(-0.62) (-0.92) (-2.68) (1.41) (-0.35) (-1.70) (0.30) (-0.54) (-0.23)
Overweight
0.17 -0.07 0.54 -0.16 0.07 -0.13 1.73 -0.48 0.22
(0.72) (-0.27) (2.01) (-0.59) (0.14) (-0.36) (1.30) (-1.30) (0.47)
Underweight
0.05 -0.05 0.01 0.09 0.42 -0.09 -0.09 -0.39 -0.02
(0.22) (-0.19) (0.03) (0.33) (0.58) (-0.16) (-0.15) (-0.88) (-0.06)
Weighting (overweight
- underweight)
0.12 -0.02 0.53 -0.25 -0.34 -0.04 1.82 -0.09 0.24
(0.45) (-0.05) (1.61) (-0.69) (-0.62) (-0.08) (1.24) (-0.17) (0.43)
Active (screening +
weighting)
-0.15 -0.42 -1.17 0.62 -0.84 -1.79 2.13 -0.49 0.10
(-0.35) (-1.06) (-2.01) (1.14) (-0.64) (-2.00) (1.14) (-0.70) (0.14)
Benchmark
0.49 0.41 2.39 -1.03 1.32 1.47 0.15 -0.26 0.07
(2.94) (2.34) (3.60) (-1.97) (1.10) (1.46) (2.04) (-0.64) (0.21)
Total Returns (active +
benchmark)
0.34 -0.01 1.22 -0.41 0.48 -0.32 2.27 -0.75 0.17
(0.68) (-0.03) (2.08) (-0.83) (0.32) (-0.30) (1.21) (-0.97) (0.21)
Value Weighted Fund
Returns
0.22 0.00 0.69 -0.16 0.82 -0.28 0.45 -0.66 -0.08
(0.46) (0.01) (1.19) (-0.33) (0.49) (-0.21) (0.44) (-0.83) (-0.11)

70

Table 19 (continued). Average yearly returns (%) of funds grouped by their benchmark, obtained using Frank Russell index
construction methodology for years 1980-1998
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Number of funds 450 637 716 172 575 16 42

Screen-In
1.77 1.55 1.00 0.34 -5.36 -1.40 2.62
(0.87) (1.88) (0.83) (0.31) (-1.86) (-1.12) (1.88)
Screen-Out
1.22 -1.83 -0.48 1.52 2.51 0.12 -1.12
(0.76) (-2.67) (-0.81) (2.41) (1.43) (0.14) (-1.60)
Screening (screen-in +
screen-out)
2.99 -0.27 0.52 1.86 -2.85 -1.28 1.50
(1.90) (-0.37) (0.48) (1.90) (-1.24) (-1.32) (0.96)
Overweight
-0.40 0.58 0.08 0.29 -0.41 0.49 0.10
(-0.55) (1.87) (0.21) (0.70) (-0.28) (0.79) (0.04)
Underweight
1.69 0.28 0.13 0.40 -0.74 -0.76 1.89
(2.02) (0.91) (0.31) (0.82) (-0.59) (-1.71) (1.68)
Weighting (overweight
- underweight)
-2.09 0.31 -0.06 -0.11 0.33 1.25 -1.80
(-1.79) (0.87) (-0.18) (-0.27) (0.21) (1.93) (-0.65)
Active (screening +
weighting)
0.90 0.04 0.46 1.75 -2.52 -0.03 -0.30
(0.58) (0.05) (0.46) (1.88) (-1.06) (-0.03) (-0.10)
Benchmark
0.03 1.36 -0.26 -1.27 -0.83 -0.24 0.13
(0.14) (3.03) (-0.57) (-2.11) (-1.65) (-0.42) (1.59)
Total Returns (active +
benchmark)
0.93 1.40 0.20 0.48 -3.35 -0.27 -0.17
(0.60) (1.83) (0.17) (0.42) (-1.35) (-0.23) (-0.06)
Value Weighted Fund
Returns
3.03 1.09 0.26 0.59 -3.68 -1.52 1.63
(1.90) (1.45) (0.20) (0.49) (-1.51) (-1.34) (1.04)

71

Table 20. Average yearly returns (%) of funds grouped by their benchmark, obtained using Frank Russell index construction
methodology for years 1999-2016
I first compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then calculate value weighted average across funds at
time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report
𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the sample standard deviation of 𝑀 𝑐 , assuming no serial
correlation in the returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 11244 1585 1745 801 761 605 232 936 1020

Screen-In
0.27 0.00 -0.63 0.00 1.56 -0.38 -1.00 1.30 -0.45
(0.67) (0.01) (-0.38) (-0.00) (1.30) (-0.74) (-1.04) (1.14) (-0.45)
Screen-Out
0.16 0.16 2.78 -0.33 -1.51 -0.24 0.71 -0.56 1.25
(0.52) (0.57) (1.34) (-0.29) (-2.39) (-0.30) (1.68) (-1.05) (2.21)
Screening (screen-in +
screen-out)
0.43 0.16 2.15 -0.34 0.05 -0.62 -0.28 0.73 0.80
(1.18) (0.35) (1.49) (-0.25) (0.04) (-0.71) (-0.34) (0.66) (0.92)
Overweight
3.64 3.33 2.91 4.75 3.99 3.43 6.76 4.38 4.52
(2.68) (2.79) (1.19) (2.24) (2.67) (2.51) (2.73) (2.86) (2.42)
Underweight
2.40 1.82 0.75 1.75 2.81 2.75 2.75 2.69 2.56
(1.91) (1.57) (0.34) (0.85) (1.96) (2.07) (1.26) (1.78) (1.48)
Weighting (overweight
- underweight)
1.24 1.51 2.16 3.01 1.18 0.68 4.01 1.69 1.95
(2.68) (2.37) (1.78) (2.36) (2.37) (1.68) (2.11) (2.36) (1.88)
Active (screening +
weighting)
1.67 1.67 4.31 2.67 1.23 0.06 3.73 2.42 2.75
(3.13) (3.17) (2.25) (1.66) (0.92) (0.08) (2.00) (1.75) (2.12)
Benchmark
7.62 6.53 0.24 6.58 10.83 11.02 6.49 8.14 6.16
(2.13) (1.90) (0.04) (1.22) (2.67) (2.49) (1.79) (2.18) (1.75)
Total Returns (active +
benchmark)
9.29 8.20 4.56 9.25 12.05 11.08 10.22 10.57 8.91
(2.47) (2.41) (0.73) (1.90) (2.62) (2.47) (2.52) (2.65) (2.19)
Value Weighted Fund
Returns
8.05 6.69 2.39 6.24 10.88 10.40 6.21 8.88 6.95
(2.21) (1.97) (0.40) (1.30) (2.40) (2.35) (1.69) (2.28) (1.78)

72

Table 20 (continued). DGTW adjusted average yearly returns (%) of funds grouped by their benchmark, obtained using Frank
Russell index construction methodology for years 1999-2016
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 101 349 425 584 172 628 31 1269

Screen-In
2.38 1.42 -1.02 -0.96 0.57 3.97 0.28 -0.76
(1.63) (1.47) (-0.81) (-0.95) (0.28) (3.27) (0.09) (-1.27)
Screen-Out
1.46 -0.70 0.43 -1.16 0.44 -1.42 0.02 0.68
(1.24) (-1.23) (0.78) (-1.90) (0.34) (-2.67) (0.01) (1.51)
Screening (screen-in +
screen-out)
3.84 0.72 -0.59 -2.12 1.01 2.55 0.30 -0.08
(2.50) (0.79) (-0.49) (-1.95) (0.61) (2.24) (0.13) (-0.15)
Overweight
2.88 3.67 3.05 2.88 4.25 4.70 6.76 6.01
(2.26) (2.69) (2.04) (1.93) (2.53) (3.73) (1.81) (3.39)
Underweight
4.06 3.63 2.34 1.83 3.29 3.88 4.61 4.24
(3.04) (2.69) (1.56) (1.24) (1.81) (3.35) (1.10) (2.90)
Weighting (overweight
- underweight)
-1.18 0.04 0.71 1.05 0.96 0.82 2.15 1.77
(-1.85) (0.10) (1.68) (2.16) (0.90) (2.31) (0.62) (2.43)
Active (screening +
weighting)
2.67 0.75 0.11 -1.08 1.97 3.37 2.45 1.69
(1.89) (0.80) (0.10) (-1.04) (1.31) (2.87) (0.88) (2.63)
Benchmark
10.67 11.71 9.78 9.51 10.47 10.67 9.46 11.91
(2.39) (2.69) (2.19) (2.37) (2.81) (2.70) (1.67) (3.07)
Total Returns (active +
benchmark)
13.33 12.46 9.90 8.43 12.44 14.04 11.91 13.60
(2.86) (2.99) (1.98) (1.85) (2.88) (3.78) (1.99) (3.25)
Value Weighted Fund
Returns
14.51 12.42 9.19 7.38 11.48 13.22 9.75 11.83
(3.08) (3.00) (1.84) (1.62) (2.59) (3.67) (1.57) (3.04)

73

Table 21. DGTW adjusted average yearly returns (%) of funds grouped by their benchmark, obtained using Frank Russell
index construction methodology for years 1999-2016
I first compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then calculate value weighted average across funds at
time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report
𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the sample standard deviation of 𝑀 𝑐 , assuming no serial
correlation in the returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 11244 1585 1745 801 761 605 232 936 1020

Screen-In
0.43 -0.03 -0.60 0.70 1.53 -0.16 -0.34 1.74 0.13
(1.69) (-0.07) (-0.67) (0.55) (2.23) (-0.35) (-0.54) (1.88) (0.21)
Screen-Out
0.04 0.05 1.36 -1.13 -0.70 0.41 0.37 -0.92 0.75
(0.19) (0.28) (1.12) (-1.49) (-1.05) (0.63) (1.40) (-2.54) (2.08)
Screening (screen-in +
screen-out)
0.47 0.02 0.76 -0.43 0.83 0.25 0.03 0.81 0.87
(1.75) (0.07) (0.67) (-0.37) (0.96) (0.41) (0.04) (0.98) (1.45)
Overweight
0.56 0.64 0.59 1.52 0.75 0.15 3.00 0.80 1.00
(3.24) (3.00) (1.10) (1.74) (3.13) (0.52) (2.63) (1.60) (2.03)
Underweight
0.09 -0.20 -0.55 0.08 0.38 0.19 -0.35 0.25 -0.08
(0.71) (-0.98) (-1.46) (0.13) (1.32) (0.90) (-0.75) (0.69) (-0.21)
Weighting (overweight
- underweight)
0.47 0.85 1.14 1.44 0.37 -0.03 3.35 0.55 1.08
(2.14) (2.49) (1.64) (1.82) (1.38) (-0.11) (2.48) (1.01) (1.70)
Active (screening +
weighting)
0.94 0.87 1.89 1.01 1.20 0.21 3.37 1.36 1.95
(3.02) (2.77) (1.45) (0.75) (1.49) (0.36) (2.26) (1.30) (2.34)
Benchmark
0.20 0.19 -1.78 1.57 0.91 0.28 -0.10 0.81 -0.16
(0.78) (0.81) (-1.45) (1.21) (1.14) (0.30) (-0.89) (1.52) (-0.60)
Total Returns (active +
benchmark)
1.14 1.06 0.11 2.58 2.11 0.50 3.27 2.17 1.79
(3.49) (2.83) (0.13) (1.46) (3.02) (0.65) (2.20) (1.84) (2.11)
Value Weighted Fund
Returns
0.67 0.21 -1.02 1.14 1.74 0.53 -0.08 1.62 0.72
(2.62) (0.57) (-1.35) (0.76) (2.38) (0.81) (-0.12) (1.70) (1.13)

74

Table 21 (continued). DGTW adjusted average yearly returns (%) of funds grouped by their benchmark, obtained using Frank
Russell index construction methodology for years 1999-2016
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 101 349 425 584 172 628 31 1269

Screen-In
2.54 1.09 -0.56 0.01 1.65 4.25 0.59 -0.15
(1.87) (1.39) (-0.67) (0.01) (0.93) (3.63) (0.30) (-0.41)
Screen-Out
1.95 0.07 0.61 -0.99 0.86 -1.56 -0.43 0.29
(2.04) (0.12) (1.33) (-1.76) (0.75) (-3.16) (-0.37) (0.72)
Screening (screen-in +
screen-out)
4.49 1.16 0.05 -0.98 2.51 2.69 0.16 0.14
(3.64) (1.42) (0.06) (-1.47) (1.84) (2.59) (0.10) (0.30)
Overweight
0.20 -0.14 0.03 -0.09 0.62 1.32 1.13 0.63
(0.49) (-0.34) (0.11) (-0.30) (0.98) (3.02) (0.46) (2.51)
Underweight
1.91 0.65 -0.17 -0.04 1.20 1.43 0.28 -0.07
(3.98) (1.89) (-0.47) (-0.12) (1.46) (3.95) (0.19) (-0.33)
Weighting (overweight
- underweight)
-1.71 -0.79 0.20 -0.04 -0.58 -0.11 0.85 0.70
(-2.98) (-2.12) (0.63) (-0.13) (-0.62) (-0.31) (0.32) (2.15)
Active (screening +
weighting)
2.78 0.37 0.25 -1.02 1.92 2.58 1.01 0.84
(2.40) (0.46) (0.35) (-1.65) (1.53) (2.49) (0.39) (1.80)
Benchmark
0.14 0.96 -0.54 1.30 1.54 2.10 -0.09 0.16
(0.28) (1.21) (-1.02) (1.78) (2.41) (2.47) (-0.37) (0.38)
Total Returns (active +
benchmark)
2.92 1.33 -0.29 0.28 3.46 4.68 0.92 1.00
(2.70) (1.17) (-0.33) (0.31) (2.60) (3.79) (0.35) (1.90)
Value Weighted Fund
Returns
4.63 2.12 -0.49 0.32 4.05 4.79 0.07 0.30
(3.88) (2.01) (-0.50) (0.31) (2.60) (3.99) (0.04) (0.60)

75

Table 22. Average yearly returns (%) of funds grouped by their benchmark for years 1980-2002
I first compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then calculate value weighted average across funds at
time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report
𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the sample standard deviation of 𝑀 𝑐 , assuming no serial
correlation in the returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Number of funds 9113 1937 2808 2187 1292 889

Screen-In
-0.05 -0.59 0.99 -0.97 2.34 -1.32
(-0.07) (-0.72) (0.95) (-0.98) (1.31) (-1.25)
Screen-Out
-0.62 -0.14 -2.83 1.94 -2.92 0.11
(-1.21) (-0.27) (-2.69) (2.30) (-2.62) (0.07)
Screening (screen-in +
screen-out)
-0.67 -0.73 -1.84 0.97 -0.57 -1.21
(-1.05) (-1.00) (-1.78) (1.25) (-0.32) (-0.75)
Overweight
5.70 5.23 5.78 5.74 4.68 3.74
(3.91) (3.82) (3.76) (4.01) (2.25) (1.55)
Underweight
5.02 4.42 4.94 -5.40 4.82 3.20
(3.72) (3.32) (3.51) (4.08) (2.33) (1.30)
Weighting (overweight
- underweight)
0.68 0.81 0.84 0.34 -0.14 0.54
(1.18) (1.17) (1.32) (0.54) (-0.17) (0.68)
Active (screening +
weighting)
0.01 0.08 -1.00 1.31 -0.71 -0.67
(0.01) (0.09) (-0.81) (1.24) (-0.38) (-0.46)
Benchmark
13.99 13.53 15.36 12.19 13.43 10.54
(4.30) (4.16) (4.38) (3.78) (2.73) (1.55)
Total Returns (active +
benchmark)
14.00 13.62 14.36 13.50 12.72 9.86
(3.87) (3.88) (3.73) (4.04) (2.22) (1.40)
Value Weighted Fund
Returns
13.33 12.81 13.52 13.16 12.86 9.32
(3.80) (3.69) (3.63) (4.07) (2.25) (1.32)

76

Table 23. DGTW adjusted average yearly returns (%) of funds grouped by their benchmark for years 1980-2002
I first compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then calculate value weighted average across funds at
time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report
𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the sample standard deviation of 𝑀 𝑐 , assuming no serial
correlation in the returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Number of funds 9113 1937 2808 2187 1292 889

Screen-In
0.07 -0.25 0.58 -0.22 1.72 -1.45
(0.14) (-0.51) (0.87) (-0.33) (1.43) (-1.45)
Screen-Out
-0.37 0.02 -2.07 1.41 -2.04 0.29
(-1.12) (0.06) (-2.82) (2.44) (-1.88) (0.21)
Screening (screen-in +
screen-out)
-0.31 -0.23 -1.48 1.19 -0.32 -1.16
(-0.77) (-0.57) (-2.27) (1.98) (-0.24) (-0.99)
Overweight
0.36 0.22 0.63 0.24 0.39 -0.19
(1.57) (0.81) (2.27) (0.70) (0.99) (-0.32)
Underweight
0.05 -0.12 -0.11 0.29 0.57 0.06
(0.27) (-0.50) (-0.46) (1.13) (1.12) (0.14)
Weighting (overweight
- underweight)
0.30 0.34 0.74 -0.05 -0.19 -0.25
(1.18) (0.94) (2.33) (-0.15) (-0.41) (-0.41)
Active (screening +
weighting)
0.00 0.11 -0.75 1.14 -0.50 -1.41
(-0.00) (0.25) (-1.12) (1.78) (-0.41) (-1.23)
Benchmark
0.70 0.35 1.92 -0.73 1.97 0.99
(3.12) (2.13) (2.91) (-1.24) (1.59) (0.65)
Total Returns (active +
benchmark)
0.70 0.46 1.17 0.41 1.47 -0.41
(1.51) (0.94) (1.99) (0.58) (1.37) (-0.31)
Value Weighted Fund
Returns
0.39 0.12 0.43 0.46 1.65 -0.16
(0.95) (0.29) (0.79) (0.76) (1.40) (-0.14)

77

Table 24. Average yearly returns (%) of funds grouped by their benchmark for years 2003-2016
I first compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then calculate value weighted average across funds at
time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report
𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the sample standard deviation of 𝑀 𝑐 , assuming no serial
correlation in the returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 10266 1386 1041 488 565 437 907 993 1146

Screen-In
-0.42 -0.28 -1.86 0.39 0.75 0.69 1.60 -0.76 -0.05
(-1.10) (-0.52) (-1.83) (0.20) (1.32) (1.11) (1.79) (-0.83) (-0.05)
Screen-Out
0.46 0.51 5.99 -0.72 -0.22 -0.36 -2.42 0.24 -0.76
(1.24) (0.95) (3.44) (-0.59) (-0.42) (-0.53) (-2.69) (0.29) (-0.95)
Screening (screen-in +
screen-out)
0.04 0.23 4.13 -0.33 0.53 0.33 -0.82 -0.52 -0.81
(0.10) (0.39) (3.07) (-0.20) (0.88) (0.36) (-0.93) (-0.61) (-1.05)
Overweight
4.26 3.98 6.88 8.90 4.02 3.83 3.97 3.96 3.74
(2.89) (3.47) (2.84) (3.91) (2.92) (2.87) (2.07) (2.08) (1.86)
Underweight
3.46 2.83 4.88 6.35 3.60 3.51 3.22 3.27 3.56
(2.77) (2.80) (2.95) (3.34) (2.93) (2.82) (2.04) (2.01) (2.10)
Weighting (overweight
- underweight)
0.80 1.15 2.00 2.56 0.42 0.32 0.76 0.69 0.18
(1.79) (2.63) (1.45) (2.21) (1.18) (0.89) (1.06) (1.04) (0.21)
Active (screening +
weighting)
0.84 1.38 6.13 2.23 0.95 0.65 -0.06 0.17 -0.63
(1.61) (2.20) (2.84) (1.22) (1.52) (0.88) (-0.06) (0.18) (-0.69)
Benchmark
11.02 10.19 9.95 19.05 12.64 13.09 9.84 9.70 10.17
(2.87) (2.81) (2.22) (3.61) (2.86) (2.70) (2.22) (2.22) (2.34)
Total Returns (active +
benchmark)
11.86 11.57 16.07 21.27 13.59 13.74 9.78 9.86 9.54
(2.89) (3.04) (2.90) (4.08) (2.94) (2.84) (2.03) (2.11) (1.97)
Value Weighted Fund
Returns
11.07 10.42 14.08 18.72 13.18 13.42 9.03 9.17 9.36
(2.85) (2.85) (2.96) (3.86) (2.94) (2.83) (2.01) (2.08) (2.07)

78

Table 24 (continued). Average yearly returns (%) of funds grouped by their benchmark for years 2003-2016
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 187 264 316 479 199 305 315 1238

Screen-In
-0.34 -0.29 -0.70 -0.15 -1.00 -0.32 0.36 0.37
(-0.56) (-0.42) (-0.76) (-0.19) (-0.90) (-0.20) (0.38) (0.21)
Screen-Out
0.45 0.67 0.15 -0.25 0.37 0.25 0.20 0.73
(0.52) (0.71) (0.20) (-0.30) (0.41) (0.45) (0.18) (0.40)
Screening (screen-in +
screen-out)
0.11 0.37 -0.55 -0.40 -0.63 -0.07 0.56 1.09
(0.16) (0.40) (-0.60) (-0.50) (-0.71) (-0.04) (0.54) (0.89)
Overweight
3.66 3.48 3.16 3.33 4.27 3.38 3.60 5.20
(2.50) (1.94) (1.89) (1.94) (2.21) (2.07) (2.18) (2.97)
Underweight
3.31 3.14 2.86 2.56 3.11 2.89 3.04 3.74
(2.47) (1.95) (1.80) (1.61) (1.86) (1.89) (2.28) (2.43)
Weighting (overweight
- underweight)
0.35 0.34 0.30 0.77 1.16 0.49 0.57 1.46
(1.16) (0.76) (0.78) (1.96) (1.63) (1.58) (0.95) (2.40)
Active (screening +
weighting)
0.46 0.72 -0.25 0.37 0.53 0.42 1.13 2.56
(0.66) (0.89) (-0.33) (0.55) (0.55) (0.28) (1.03) (1.87)
Benchmark
13.01 11.50 11.32 9.71 10.64 10.32 10.41 12.70
(2.34) (1.97) (1.89) (1.80) (1.99) (1.91) (2.27) (3.30)
Total Returns (active +
benchmark)
13.46 12.21 11.07 10.08 11.17 10.74 11.54 15.26
(2.41) (2.10) (1.88) (1.84) (2.00) (1.90) (2.38) (3.44)
Value Weighted Fund
Returns
13.11 11.87 10.77 9.31 10.01 10.25 10.97 13.79
(2.40) (2.11) (1.85) (1.74) (1.88) (1.85) (2.41) (3.26)

79

Table 25. DGTW adjusted average yearly returns (%) of funds grouped by their benchmark for years 2003-2016
I first compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then calculate value weighted average across funds at
time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report
𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the sample standard deviation of 𝑀 𝑐 , assuming no serial
correlation in the returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 10266 1386 1041 488 565 437 907 993 1146

Screen-In
-0.06 0.00 -0.48 0.76 0.79 0.89 1.40 -0.43 0.25
(-0.23) (0.01) (-0.64) (0.48) (1.51) (1.68) (2.40) (-0.57) (0.40)
Screen-Out
0.24 0.22 3.44 -1.27 0.10 -0.08 -1.70 0.27 -0.66
(0.89) (0.56) (2.92) (-1.34) (0.20) (-0.16) (-2.66) (0.35) (-1.16)
Screening (screen-in +
screen-out)
0.19 0.23 2.96 -0.52 0.89 0.81 -0.30 -0.16 -0.40
(0.59) (0.46) (2.92) (-0.38) (1.70) (1.18) (-0.43) (-0.27) (-0.69)
Overweight
0.33 0.44 0.16 1.22 0.38 0.37 0.35 0.19 0.19
(2.70) (2.55) (0.33) (1.68) (2.06) (1.84) (1.25) (0.82) (0.54)
Underweight
0.05 -0.15 -0.36 0.40 0.30 0.28 0.19 0.09 0.23
(0.38) (-0.90) (-0.98) (0.73) (1.80) (1.46) (0.73) (0.33) (0.68)
Weighting (overweight
- underweight)
0.29 0.60 0.52 0.83 0.07 0.10 0.17 0.10 -0.03
(1.50) (2.28) (0.75) (1.36) (0.40) (0.42) (0.42) (0.29) (-0.07)
Active (screening +
weighting)
0.47 0.82 3.49 0.31 0.96 0.90 -0.13 -0.06 -0.43
(1.49) (1.68) (2.99) (0.20) (1.94) (1.47) (-0.19) (-0.10) (-0.76)
Benchmark
0.24 0.22 -3.82 2.82 0.22 0.21 0.76 0.52 1.03
(0.75) (0.78) (-3.71) (2.91) (0.36) (0.24) (1.55) (0.84) (1.54)
Total Returns (active +
benchmark)
0.71 1.04 -0.33 3.13 1.19 1.11 0.63 0.46 0.60
(2.86) (2.57) (-0.41) (1.94) (2.12) (1.77) (1.15) (0.78) (0.83)
Value Weighted Fund
Returns
0.42 0.45 -0.86 2.31 1.12 1.01 0.46 0.36 0.63
(1.79) (1.07) (-1.37) (1.66) (2.06) (1.73) (0.84) (0.57) (0.94)

80

Table 25 (continued). DGTW adjusted average yearly returns (%) of funds grouped by their benchmark for years 2003-2016
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 187 264 316 479 199 305 315 1238

Screen-In
-0.43 -0.46 -0.40 0.33 -0.81 -0.20 0.23 0.00
(-0.69) (-0.65) (-0.48) (0.47) (-0.76) (-0.13) (0.28) (-0.00)
Screen-Out
1.02 0.96 -0.05 -0.07 0.91 0.26 0.59 1.01
(1.37) (1.21) (-0.08) (-0.09) (0.93) (0.45) (0.54) (0.56)
Screening (screen-in +
screen-out)
0.59 0.50 -0.45 0.25 0.10 0.06 0.82 1.01
(0.96) (0.73) (-0.68) (0.45) (0.14) (0.04) (0.86) (1.00)
Overweight
0.14 0.33 0.40 0.37 0.57 0.17 0.42 0.64
(0.60) (1.13) (1.53) (1.59) (1.18) (0.41) (1.24) (2.13)
Underweight
0.18 0.11 0.16 -0.04 0.03 0.04 0.16 -0.07
(0.86) (0.40) (0.44) (-0.11) (0.07) (0.10) (0.58) (-0.24)
Weighting (overweight
- underweight)
-0.04 0.22 0.24 0.40 0.54 0.13 0.26 0.70
(-0.15) (0.79) (0.73) (1.94) (0.87) (0.60) (0.77) (1.96)
Active (screening +
weighting)
0.55 0.72 -0.21 0.66 0.64 0.19 1.07 1.72
(0.88) (1.07) (-0.35) (1.16) (0.76) (0.13) (1.14) (1.66)
Benchmark
0.07 0.43 1.24 0.06 0.35 0.37 -0.01 0.11
(0.07) (0.39) (1.27) (0.06) (0.50) (0.49) (-0.02) (0.26)
Total Returns (active +
benchmark)
0.62 1.15 1.03 0.72 0.99 0.55 1.06 1.82
(0.83) (1.32) (1.03) (0.85) (1.03) (0.32) (1.08) (1.99)
Value Weighted Fund
Returns
0.66 0.93 0.80 0.32 0.45 0.43 0.80 1.12
(0.91) (1.12) (0.71) (0.34) (0.52) (0.24) (0.85) (1.27)

81

Table 26. Average yearly returns (%) of funds grouped by their benchmark without months of extreme returns
I remove periods of … as months of extreme returns. I compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then
calculate value weighted average across funds at time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure
and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report 𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the
sample standard deviation of 𝑀 𝑐 , assuming no serial correlation in the returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 16163 2402 2868 2229 1393 931 905 990 1144

Screen-In
-0.20 -0.49 0.57 -0.69 1.66 -0.15 1.50 -0.94 -0.07
(-0.41) (-0.90) (0.62) (-0.76) (1.91) (-0.29) (1.67) (-1.03) (-0.07)
Screen-Out
-0.14 0.19 -1.61 1.68 -1.40 -0.14 -2.39 0.38 -0.81
(-0.41) (0.48) (-1.67) (2.21) (-2.38) (-0.19) (-2.63) (0.46) (-1.01)
Screening (screen-in +
screen-out)
-0.34 -0.30 -1.04 0.99 0.27 -0.29 -0.89 -0.56 -0.88
(-0.79) (-0.60) (-1.12) (1.41) (0.30) (-0.37) (-1.01) (-0.65) (-1.14)
Overweight
5.55 5.11 6.31 6.46 4.45 3.92 4.43 4.32 4.22
(5.46) (5.64) (4.82) (5.37) (3.67) (3.18) (2.36) (2.30) (2.14)
Underweight
4.70 4.03 5.18 5.74 4.26 3.55 3.44 3.45 3.87
(5.05) (4.58) (4.37) (5.10) (3.68) (2.96) (2.19) (2.11) (2.30)
Weighting (overweight
- underweight)
0.85 1.07 1.13 0.72 0.19 0.37 0.99 0.87 0.34
(2.19) (2.34) (1.97) (1.27) (0.45) (1.00) (1.44) (1.36) (0.40)
Active (screening +
weighting)
0.51 0.77 0.09 1.71 0.46 0.08 0.10 0.31 -0.54
(0.89) (1.31) (0.08) (1.84) (0.50) (0.12) (0.10) (0.34) (-0.58)
Benchmark
13.57 12.89 15.28 13.46 13.15 12.71 10.66 10.30 11.16
(5.62) (5.46) (5.10) (4.86) (4.02) (3.21) (2.42) (2.36) (2.61)
Total Returns (active +
benchmark)
14.08 13.66 15.37 15.17 13.61 12.79 10.76 10.61 10.62
(5.36) (5.46) (4.68) (5.38) (3.76) (3.19) (2.26) (2.28) (2.23)
Value Weighted Fund
Returns
13.23 12.58 14.24 14.44 13.42 12.42 9.77 9.74 10.28
(5.21) (5.09) (4.50) (5.28) (3.77) (3.13) (2.19) (2.21) (2.30)

82

Table 26 (continued). Average yearly returns (%) of funds grouped by their benchmark without months of extreme returns
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 187 263 316 478 199 305 315 1238

Screen-In
-0.26 -0.24 -0.66 -0.14 -0.73 -0.61 0.32 0.37
(-0.43) (-0.34) (-0.71) (-0.17) (-0.68) (-0.38) (0.34) (0.21)
Screen-Out
0.54 0.93 0.29 -0.26 0.06 0.29 0.20 0.73
(0.63) (1.01) (0.40) (-0.30) (0.07) (0.52) (0.18) (0.40)
Screening (screen-in +
screen-out)
0.28 0.69 -0.37 -0.40 -0.67 -0.32 0.52 1.09
(0.42) (0.76) (-0.40) (-0.49) (-0.76) (-0.21) (0.49) (0.89)
Overweight
3.89 3.71 3.48 3.72 4.70 3.58 3.91 5.20
(2.67) (2.07) (2.11) (2.21) (2.48) (2.19) (2.39) (2.97)
Underweight
3.56 3.42 3.20 2.99 3.45 3.10 3.26 3.74
(2.69) (2.14) (2.05) (1.94) (2.09) (2.03) (2.46) 2.43)
Weighting (overweight
- underweight)
0.33 0.29 0.28 0.73 1.25 0.48 0.65 1.46
(1.10) (0.65) (0.71) (1.86) (1.76) (1.55) (1.09) (2.40)
Active (screening +
weighting)
0.61 0.98 -0.09 0.33 0.58 0.17 1.17 2.56
(0.90) (1.27) (-0.12) (0.49) (0.60) (0.11) (1.07) (1.87)
Benchmark
13.72 12.04 12.32 11.06 11.76 11.23 11.23 12.70
(2.47) (2.05) (2.07) (2.10) (2.23) (2.09) (2.46) (3.30)
Total Returns (active +
benchmark)
14.33 13.02 12.23 11.39 12.34 11.40 12.40 15.26
(2.58) (2.25) (2.10) (2.13) (2.25) (2.01) (2.58) (3.44)
Value Weighted Fund
Returns
14.00 12.72 11.96 10.66 11.09 10.91 11.75 13.79
(2.58) (2.27) (2.09) (2.05) (2.11) (1.96) (2.59) (3.26)

83

Table 27. DGTW adjusted average yearly returns (%) of funds grouped by their benchmark without months of extreme returns
I remove periods of … as months of extreme returns. I compute each measure of interest 𝑀 𝑡 𝑓 for a fund f at time t following methodology described above. I then
calculate value weighted average across funds at time t for each category of funds, c, 𝑀 𝑡 𝑐 =
1
𝐹 ∑ 𝑀 𝑡 𝑓 𝑓 ∈𝑐 . Finally, I compute a time-series average for each measure
and each category of funds, 𝑀 𝑐 =
1
𝑇 ∑ 𝑀 𝑡 𝑐 𝑡 . I report 𝑀 𝑐 for each category of funds, as well as the t-statistic for the measure. The standard error is calculated as the
sample standard deviation of 𝑀 𝑐 , assuming no serial correlation in the returns.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Number of funds 16163 2402 2868 2229 1393 931 905 990 1144

Screen-In
0.03 -0.16 0.41 0.00 1.32 -0.10 1.38 -0.50 0.28
(0.09) (-0.47) (0.70) (0.00) (2.13) (-0.20) (2.35) (-0.67) (0.43)
Screen-Out
-0.15 0.09 -1.40 1.04 -0.80 0.19 -1.71 0.32 -0.72
(-0.63) (0.33) (-2.10) (1.99) (-1.41) (0.32) (-2.65) (0.40) (-1.27)
Screening (screen-in +
screen-out)
-0.12 -0.07 -0.99 1.04 0.53 0.09 -0.33 -0.19 -0.44
(-0.44) (-0.23) (-1.67) (1.89) (0.81) (0.16) (-0.47) (-0.31) (-0.77)
Overweight
0.36 0.34 0.60 0.35 0.33 0.05 0.40 0.20 0.23
(2.46) (1.89) (2.42) (1.12) (1.63) (0.21) (1.41) (0.84) (0.64)
Underweight
0.04 -0.15 -0.14 0.28 0.44 0.18 0.13 0.02 0.21
(0.30) (-0.90) (-0.66) (1.22) (1.73) (0.91) (0.51) (0.06) (0.64)
Weighting (overweight
- underweight)
0.32 0.49 0.74 0.06 -0.11 -0.13 0.27 0.18 0.01
(1.83) (1.97) (2.56) (0.22) (-0.51) (-0.50) (0.69) (0.54) (0.03)
Active (screening +
weighting)
0.20 0.42 -0.24 1.11 0.41 -0.04 -0.06 -0.01 -0.43
(0.70) (1.26) (-0.40) (1.87) (0.67) (-0.07) (-0.08) (-0.01) (-0.75)
Benchmark
0.50 0.30 1.30 -0.37 0.82 0.31 0.69 0.37 1.06
(2.75) (2.04) (2.17) (-0.71) (1.26) (0.39) (1.42) (0.62) (1.58)
Total Returns (active +
benchmark)
0.71 0.72 1.05 0.73 1.23 0.27 0.64 0.37 0.63
(2.37) (2.11) (2.01) (1.13) (2.13) (0.45) (1.16) (0.63) (0.88)
Value Weighted Fund
Returns
0.38 0.23 0.31 0.67 1.34 0.40 0.37 0.19 0.62
(1.42) (0.75) (0.64) (1.21) (2.16) (0.74) (0.68) (0.30) (0.92)

84

Table 27 (continued). DGTW adjusted average yearly returns (%) of funds grouped by their benchmark without months of
extreme returns
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Number of funds 187 263 316 478 199 305 315 1238

Screen-In
-0.33 -0.43 -0.50 0.26 -0.48 -0.34 0.22 0.00
(-0.52) (-0.60) (-0.59) (0.38) (-0.46) (-0.22) (0.27) (-0.00)
Screen-Out
1.03 1.12 0.05 -0.10 0.54 0.29 0.58 1.01
(1.38) (1.43) (0.09) (-0.13) (0.59) (0.50) (0.53) (0.56)
Screening (screen-in +
screen-out)
0.71 0.69 -0.45 0.15 0.06 -0.05 0.80 1.01
(1.16) (1.04) (-0.68) (0.28) (0.09) (-0.03) (0.84) (1.00)
Overweight
0.12 0.26 0.42 0.40 0.63 0.12 0.40 0.64
(0.51) (0.92) (1.61) (1.72) (1.30) (0.30) (1.17) (2.13)
Underweight
0.17 0.06 0.13 0.03 0.03 0.02 0.12 -0.07
(0.80) (0.23) (0.35) (0.08) (0.07) (0.06) (0.44) (-0.24)
Weighting (overweight
- underweight)
-0.05 0.20 0.29 0.37 0.60 0.10 0.28 0.70
(-0.19) (0.71) (0.91) (1.80) (0.96) (0.47) (0.83) (1.96)
Active (screening +
weighting)
0.66 0.89 -0.16 0.53 0.66 0.05 1.08 1.72
(1.06) (1.35) (-0.26) (0.95) (0.78) (0.03) (1.14) (1.66)
Benchmark
-0.16 0.10 1.16 0.29 0.42 0.34 -0.10 0.11
(-0.18) (0.09) (1.18) (0.30) (0.60) (0.44) (-0.15) (0.26)
Total Returns (active +
benchmark)
0.50 0.99 1.01 0.81 1.09 0.38 0.98 1.82
(0.67) (1.14) (0.99) (0.96) (1.13) (0.22) (0.99) (1.99)
Value Weighted Fund
Returns
0.55 0.79 0.71 0.44 0.49 0.29 0.70 1.12
(0.76) (0.95) (0.63) (0.47) (0.56) (0.16) (0.74) (1.27)

85

Table 28. Average yearly gross returns (%) of quintile portfolios of stocks sorted based on the number of times a stock has been
screened-out across all the funds, grouped by the fund’s benchmark
The table shows returns of value weighted quintile portfolios of stocks. The first quintile consists of stocks that were screened-out the least, and the fifth quintile
consists of stocks that were screened-out the most. 5-1 shows the return of a long-short portfolio, buying stocks that have been screened-out the most and selling
the stocks that have been screen-out the least. The numbers show the time series average of those portfolio returns. T-statistics are calculated using standard errors
of time series average and displayed in brackets.
Quintile Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
1
12.59 6.42 0.54 10.79 9.93 11.03 7.90 8.59 9.76
(4.43) (1.80) (0.08) (1.85) (2.18) (2.61) (1.67) (1.87) (2.26)
2
14.55 8.74 3.85 13.62 10.55 11.27 8.41 9.31 9.06
(5.00) (2.27) (0.62) (2.31) (2.42) (2.74) (1.48) (1.84) (1.79)
3
9.49 10.89 6.30 12.17 10.80 10.64 11.17 11.78 11.37
(3.07) (2.89) (0.83) (1.77) (2.44) (2.62) (1.93) (2.00) (2.14)
4
12.45 12.87 11.58 11.93 12.13 11.97 9.17 9.89 11.54
(4.72) (3.03) (2.15) (1.42) (2.68) (2.88) (1.48) (1.50) (2.03)
5
13.29 3.73 -1.25 2.02 8.49 9.47 8.64 9.81 8.72
(5.32) (0.89) (-0.21) (0.25) (1.97) (2.05) (1.38) (1.54) (1.47)
5-1
0.69 -2.68 -1.79 -8.77 -1.43 -1.56 0.74 1.22 -1.04
(0.50) (-1.57) (-0.35) (-2.06) (-0.68) (-0.83) (0.27) (0.38) (-0.36)

86

Table 28 (continued). Average yearly gross returns (%) of quintile portfolios of stocks sorted based on the number of times a
stock has been screened-out across all the funds, grouped by the fund’s benchmark
Quintile Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
1
6.37 10.45 10.12 8.16 75.92 9.89 10.10 10.18
(0.83) (1.82) (1.69) (1.48) (3.40) (1.86) (1.08) (2.41)
2
5.76 11.63 11.09 9.55 78.82 10.28 8.08 13.84
(0.80) (1.95) (1.78) (1.73) (2.65) (1.77) (0.78) (2.78)
3
3.28 11.25 13.30 12.82 86.54 9.16 13.77 13.37
(0.43) (1.85) (2.14) (2.34) (2.22) (1.54) (1.22) (2.54)
4
7.86 12.67 11.54 11.20 95.14 11.08 7.44 11.38
(1.03) (2.02) (1.77) (1.94) (2.15) (1.88) (0.64) (1.89)
5
2.52 8.00 9.73 7.94 94.16 7.38 11.14 8.47
(0.30) (1.26) (1.38) (1.36) (2.47) (1.08) (0.91) (1.37)
5-1
-3.85 -2.45 -0.39 -0.23 18.24 -2.51 1.04 -1.71
(-1.25) (-1.04) (-0.11) (-0.07) (1.02) (-0.73) (0.25) (-0.50)

87

Table 29. Average yearly excess (DGTW) returns (%) of quintile portfolios of stocks sorted based on the number of times a stock
has been screened-out across all the funds, grouped by the fund’s benchmark
The table shows returns of value weighted quintile portfolios of stocks, adjusting the vector of stock returns following Daniel et al. (1997). The first quintile consists
of stocks that were screened-out the least, and the fifth quintile consists of stocks that were screened-out the most. 5-1 shows the return of a long-short portfolio,
buying stocks that have been screened-out the most and selling the stocks that have been screen-out the least. The numbers show the time series average of those
portfolio returns. T-statistics are calculated using standard errors of time series average and displayed in brackets.
Quintile Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
1
0.39 0.21 -2.55 4.14 0.34 0.23 0.72 0.42 1.10
(0.57) (0.44) (-1.59) (2.03) (0.30) (0.18) (1.43) (0.49) (1.19)
2
1.60 1.70 -0.56 6.17 0.45 0.75 0.65 1.63 0.53
(2.10) (1.91) (-0.18) (2.58) (0.45) (0.63) (0.67) (1.49) (0.40)
3
-1.35 2.47 2.27 3.77 0.13 0.43 2.43 2.93 1.83
(-1.34) (2.45) (0.73) (1.56) (0.12) (0.39) (2.07) (1.99) (1.61)
4
-0.29 3.68 5.20 3.03 1.69 0.48 0.31 0.06 1.42
(-0.56) (2.74) (1.35) (1.01) (1.13) (0.35) (0.25) (0.03) (1.16)
5
1.05 -1.96 -1.55 -1.83 -1.59 -0.91 -0.34 0.47 -1.70
(2.19) (-1.75) (-0.49) (-0.73) (-1.15) (-0.64) (-0.22) (0.30) (-1.20)
5-1
0.66 -2.17 0.99 -5.97 -1.92 -1.13 -1.06 0.05 -2.80
(0.72) (-1.60) (0.28) (-1.77) (-0.97) (-0.65) (-0.62) (0.02) (-1.50)

88

Table 29 (continued). Average yearly excess (DGTW) returns (%) of quintile portfolios of stocks sorted based on the number of
times a stock has been screened-out across all the funds, grouped by the fund’s benchmark
Quintile Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
1
0.41 -0.30 0.88 -0.48 -0.70 1.14 1.64 -0.02
(0.27) (-0.20) (0.51) (-0.27) (-0.11) (0.68) (1.71) (-0.05)
2
0.50 0.82 1.11 -0.28 -1.90 1.38 -0.96 0.96
(0.31) (0.59) (0.81) (-0.18) (-0.31) (1.10) (-0.65) (1.00)
3
-1.70 0.74 3.10 2.96 5.77 -0.17 4.79 -0.15
(-1.02) (0.55) (2.52) (2.34) (0.49) (-0.13) (2.12) (-0.15)
4
3.77 2.35 2.18 0.33 14.72 2.08 -1.58 -1.27
(2.10) (1.72) (1.45) (0.20) (0.69) (1.46) (-0.80) (-1.04)
5
-1.71 -2.37 0.32 -2.50 9.46 -1.19 1.83 -3.62
(-0.73) (-1.55) (0.14) (-1.52) (0.70) (-0.55) (0.53) (-1.67)
5-1
-2.12 -2.07 -0.56 -2.02 10.16 -2.33 0.19 -3.59
(-0.75) (-0.97) (-0.17) (-0.83) (0.75) (-0.76) (0.05) (-1.62)

89

Table 30. Average yearly gross returns (%) of quintile portfolios of stocks sorted based on the number of times a stock has been
screened-in across all the funds, grouped by the fund’s benchmark
The table shows returns of value weighted quintile portfolios of stocks. The first quintile consists of stocks that were screened-in the least, and the fifth quintile
consists of stocks that were screened-in the most. 5-1 shows the return of a long-short portfolio, buying stocks that have been screened-in the most and selling the
stocks that have been screen-in the least. The numbers show the time series average of those portfolio returns. T-statistics are calculated using standard errors of
time series average and displayed in brackets.
Quintile Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
1
12.37 12.85 10.08 10.36 10.76 14.66 9.62 -8.68 8.41
(4.90) (3.47) (2.76) (2.33) (3.60) (3.29) (1.40) (-0.76) (1.47)
2
12.92 10.25 10.63 10.69 13.16 12.94 8.32 1.36 8.32
(4.78) (2.74) (2.89) (2.37) (4.33) (2.28) (1.16) (0.11) (1.36)
3
12.37 10.74 11.55 9.47 13.13 13.57 7.38 -10.56 4.47
(4.48) (2.90) (3.05) (2.25) (3.99) (2.59) (1.00) (-0.80) (0.79)
4
13.21 10.17 10.29 12.09 16.37 13.89 8.77 -3.80 6.80
(4.77) (2.83) (2.82) (2.91) (4.86) (2.68) (1.25) (-0.30) (1.19)
5
12.21 10.14 9.83 14.35 14.41 13.92 5.18 -0.58 5.57
(4.69) (2.75) (2.77) (3.91) (3.68) (2.34) (0.71) (-0.05) (0.99)
5-1
-0.16 -2.71 -0.25 3.99 3.66 -0.74 -4.44 8.10 -2.84
(-0.09) (-1.27) (-0.12) (1.45) (1.43) (-0.19) (-1.52) (2.10) (-1.31)

90

Table 30 (continued). Average yearly gross returns (%) of quintile portfolios of stocks sorted based on the number of times a
stock has been screened-in across all the funds, grouped by the fund’s benchmark
Quintile Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
1
-10.08 -8.19 5.08 6.93 - - 20.11 22.16
(-0.74) (-0.51) (0.56) (0.59) - - (1.16) (2.86)
2
-12.79 -2.64 6.62 7.00 - - 31.24 23.29
(-0.94) (-0.18) (0.66) (0.54) - - (2.25) (2.40)
3
-12.27 2.76 6.78 5.87 - - 8.28 20.06
(-0.87) (0.18) (0.64) (0.42) - - (0.48) (2.45)
4
-11.05 4.88 5.18 8.77 - - 7.11 21.54
(-0.89) (0.25) (0.49) (0.68) - - (0.32) (2.65)
5
-12.09 2.26 2.52 15.77 - - 15.82 23.89
(-0.95) (0.13) (0.22) (1.21) - - (0.65) (3.27)
5-1
-2.02 10.45 -2.56 8.84 - - -4.29 1.73
(-0.28) (1.84) (-0.62) (1.21) - - (-0.40) (0.31)

91

Table 31. Average yearly excess (DGTW) returns (%) of quintile portfolios of stocks sorted based on the number of times a stock
has been screened-in across all the funds, grouped by the fund’s benchmark
The table shows returns of value weighted quintile portfolios of stocks, adjusting the vector of stock returns following Daniel et al. (1997). The first quintile consists
of stocks that were screened-in the least, and the fifth quintile consists of stocks that were screened-in the most. 5-1 shows the return of a long-short portfolio,
buying stocks that have been screened-in the most and selling the stocks that have been screen-in the least. The numbers show the time series average of those
portfolio returns. T-statistics are calculated using standard errors of time series average and displayed in brackets.
Quintile Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
1
-0.43 0.81 -1.10 -0.87 -0.26 1.30 1.70 1.08 0.70
(-0.42) (1.05) (-1.17) (-0.70) (-0.31) (1.56) (1.17) (0.40) (0.67)
2
1.15 -0.92 -0.26 -0.62 0.19 0.03 -0.84 7.19 0.04
(1.25) (-0.99) (-0.30) (-0.45) (0.19) (0.01) (-0.63) (1.66) (0.03)
3
0.06 -0.40 0.47 -1.07 -0.18 -0.50 -1.35 -4.75 -1.60
(0.09) (-0.48) (0.54) (-0.81) (-0.21) (-0.31) (-0.90) (-0.97) (-1.14)
4
0.37 -0.54 -0.75 1.43 2.33 0.83 -0.15 2.95 -0.40
(0.69) (-0.84) (-1.23) (1.46) (2.66) (0.60) (-0.10) (1.01) (-0.37)
5
0.17 -0.17 -0.13 1.70 0.85 0.29 -1.53 2.18 -1.33
(0.48) (-0.23) (-0.21) (2.52) (0.96) (0.24) (-0.74) (1.34) (-1.35)
5-1
0.60 -0.98 0.97 2.57 1.11 -1.01 -3.23 1.10 -2.02
(0.52) (-0.89) (0.81) (1.68) (0.81) (-0.71) (-1.31) (0.42) (-1.37)

92

Table 31 (continued). Average yearly excess (DGTW) returns (%) of quintile portfolios of stocks sorted based on the number of
times a stock has been screened-in across all the funds, grouped by the fund’s benchmark
Quintile Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
1
0.51 -6.31 1.79 -0.12 - - -2.03 3.40
(0.12) (-1.97) (0.95) (-0.04) - - (-0.60) (1.29)
2
1.28 -2.27 1.11 -1.44 - - 7.52 2.67
(0.32) (-0.68) (0.41) (-0.50) - - (0.66) (0.71)
3
-1.01 4.04 1.81 -3.29 - - -14.90 0.65
(-0.29) (1.34) (0.64) (-1.00) - - (-1.52) (0.20)
4
0.75 5.64 3.13 -0.01 - - -10.28 0.64
(0.25) (0.74) (1.22) (-0.00) - - (-1.25) (0.23)
5
-0.53 3.02 -0.75 3.90 - - -2.01 5.81
(-0.20) (0.98) (-0.34) (1.26) - - (-0.22) (1.70)
5-1
-1.04 9.33 -2.53 4.01 - - 0.02 2.41
(-0.22) (2.38) (-0.90) (0.77) - - (0.00) (0.56)

93

Table 32. Average yearly gross returns (%) of funds grouped by their benchmark and screening intensity
The table displays the top and bottom decile portfolio returns, sorted based on the screening intensity for each category of funds. Fund returns in each portfolio are
weighted by their total assets. The bottom panel shows the results for the top-bottom portfolio. Finally, t-statistics are reported in brackets.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Top decile screening intensity
Screening
-0.79 -0.18 -0.93 1.82 -0.15 -0.07 -5.26 -2.44 -2.68
(-0.95) (-0.16) (-0.68) (1.61) (-0.15) (-0.09) (-2.02) (-1.06) (-1.38)
Weighting
0.94 1.07 1.06 0.86 0.39 0.62 0.68 0.72 -0.08
(2.34) (2.46) (1.68) (1.44) (0.71) (1.31) (0.89) (1.28) (-0.12)
Total Returns (screen +
weight + benchmark)
13.48 12.86 14.80 15.73 13.24 12.68 5.06 7.97 7.42
(4.45) (4.35) (3.95) (4.88) (3.60) (3.27) (0.83) (1.48) (1.34)
Bottom decile screening intensity
Screening
-0.36 0.02 -0.97 0.35 0.36 0.11 -0.38 -0.48 -0.61
(-0.94) (0.07) (-1.28) (0.45) (0.54) (0.18) (-0.74) (-0.47) (-0.79)
Weighting
0.61 0.27 0.82 0.87 0.25 0.10 0.29 0.44 0.17
(1.58) (1.01) (1.34) (1.28) (0.69) (0.33) (0.84) (0.65) (0.20)
Total Returns (screen +
weight + benchmark)
12.85 12.26 14.52 14.27 13.62 12.33 9.55 9.65 9.74
(5.02) (4.83) (4.41) (4.86) (4.00) (3.23) (1.95) (2.08) (2.07)
Top - bottom
Screening
-0.43 -0.20 0.04 1.47 -0.51 -0.18 -4.88 -1.96 -2.07
(-0.53) (-0.18) (0.03) (1.22) (-0.57) (-0.23) (-1.84) (-0.79) (-0.97)
Weighting
0.33 0.80 0.24 -0.01 0.14 0.53 0.39 0.28 -0.25
(0.91) (1.83) (0.44) (-0.02) (0.28) (1.15) (0.60) (0.44) (-0.42)
Total Returns (screen +
weight + benchmark)
0.63 0.60 0.28 1.46 -0.37 0.35 -4.50 -1.68 -2.32
(0.66) (0.54) (0.21) (1.18) (-0.39) (0.46) (-1.62) (-0.68) (-1.09)

94

Table 32 (continued). Average yearly gross returns (%) of funds grouped by their benchmark and screening intensity
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Top decile screening intensity
Screening
-5.12 -1.06 0.11 -1.42 -1.03 -4.00 0.75 3.10
(-1.53) (-0.44) (0.07) (-0.69) (-0.34) (-1.96) (0.27) (0.96)
Weighting
4.96 -0.81 1.10 1.01 3.68 0.87 0.14 1.87
(1.90) (-0.96) (1.73) (1.75) (1.10) (1.15) (0.09) (1.53)
Total Returns (screen +
weight + benchmark)
7.18 9.17 12.53 9.30 8.16 7.19 9.77 15.82
(0.86) (1.59) (2.12) (1.63) (0.61) (1.22) (1.12) (2.79)
Bottom decile screening intensity
Screening
1.10 0.53 -0.85 0.71 -2.80 -0.54 2.74 -1.02
(0.86) (0.62) (-0.91) (0.94) (-1.81) (-0.53) (1.37) (-1.24)
Weighting
-0.78 -0.21 0.51 0.36 1.39 -0.16 2.25 1.16
(-1.22) (-0.40) (1.07) (0.99) (1.57) (-0.23) (2.46) (1.96)
Total Returns (screen +
weight + benchmark)
7.66 11.35 10.99 10.79 4.09 9.62 13.86 11.00
(1.07) (1.93) (1.84) (2.00) (0.30) (1.79) (1.70) (2.41)
Top - bottom
Screening
-6.21 -1.59 0.96 -2.14 1.77 -3.46 -1.98 4.11
(-1.78) (-0.67) (0.60) (-1.03) (0.59) (-1.38) (-0.87) (1.29)
Weighting
5.73 -0.59 0.59 0.64 2.30 1.03 -2.10 0.70
(2.13) (-0.65) (0.86) (1.12) (0.68) (1.08) (-1.11) (0.58)
Total Returns (screen +
weight + benchmark)
-0.48 -2.19 1.54 -1.49 4.07 -2.43 -4.09 4.82
(-0.14) (-0.98) (1.04) (-0.70) (1.38) (-1.05) (-1.47) (1.47)

95

Table 33. DGTW adjusted average yearly gross returns (%) of funds grouped by their benchmark and screening intensity
The table displays the top and bottom decile portfolio returns, sorted based on the screening intensity for each category of funds. Fund returns in each portfolio are
weighted by their total assets. The bottom panel shows the results for the top-bottom portfolio. Finally, t-statistics are reported in brackets.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Top decile screening intensity
Screening
-0.52 0.00 -0.98 1.58 0.17 0.06 -3.11 -1.71 -2.21
(-1.05) (-0.00) (-1.18) (2.13) (0.24) (0.10) (-1.62) (-0.99) (-1.47)
Weighting
0.56 0.67 0.92 0.35 0.22 0.13 0.06 0.32 -0.44
(3.19) (2.44) (2.85) (1.02) (0.75) (0.34) (0.12) (0.78) (-1.05)
Total Returns (screen +
weight + benchmark)
0.77 0.97 1.13 1.65 1.42 0.69 -2.06 -0.88 -1.62
(1.65) (1.54) (1.54) (2.18) (2.42) (0.99) (-1.11) (-0.57) (-1.14)
Bottom decile screening intensity
Screening
-0.17 0.00 -0.66 0.48 0.24 0.31 -0.15 0.24 -0.11
(-0.72) (0.01) (-1.36) (0.85) (0.45) (0.67) (-0.36) (0.36) (-0.18)
Weighting
0.27 0.14 0.50 0.21 0.15 -0.19 0.17 -0.26 -0.07
(1.24) (0.84) (1.30) (0.52) (0.49) (-0.81) (0.61) (-0.60) (-0.13)
Total Returns (screen +
weight + benchmark)
0.50 0.45 1.03 0.41 1.41 0.62 1.01 0.49 0.85
(1.72) (1.49) (1.79) (0.54) (2.04) (0.68) (2.18) (0.56) (0.82)
Top - bottom
Screening
-0.35 -0.01 -0.32 1.10 -0.07 -0.25 -2.96 -1.95 -2.11
(-0.72) (-0.01) (-0.44) (1.54) (-0.11) (-0.38) (-1.53) (-1.07) (-1.33)
Weighting
0.29 0.53 0.42 0.14 0.08 0.32 -0.11 0.58 -0.37
(1.41) (1.88) (1.13) (0.30) (0.25) (0.83) (-0.20) (1.02) (-0.67)
Total Returns (screen +
weight + benchmark)
0.27 0.52 0.10 1.23 0.00 0.07 -3.07 -1.37 -2.47
(0.61) (0.79) (0.15) (1.81) (0.01) (0.12) (-1.50) (-0.75) (-1.44)

96

Table 33 (continued). DGTW adjusted average yearly gross returns (%) of funds grouped by their benchmark and screening
intensity
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Top decile screening intensity
Screening
-1.15 0.17 0.09 -0.78 -4.22 -3.05 2.07 3.12
(-0.44) (0.08) (0.07) (-0.46) (-1.85) (-1.71) (0.90) (1.17)
Weighting
3.41 -1.44 0.81 0.54 0.02 0.06 -0.12 1.48
(1.47) (-1.92) (1.32) (1.26) (0.01) (0.10) (-0.11) (1.57)
Total Returns (screen +
weight + benchmark)
1.06 -0.83 2.13 -0.17 -2.23 -2.63 1.01 4.50
(0.40) (-0.50) (2.03) (-0.11) (-0.80) (-1.37) (0.43) (1.67)
Bottom decile screening intensity
Screening
1.21 0.54 -0.59 1.04 -2.48 -0.03 3.21 -0.76
(1.09) (0.82) (-0.85) (1.66) (-2.13) (-0.03) (1.91) (-1.04)
Weighting
-1.06 -0.26 0.29 0.22 1.32 -0.59 1.74 0.25
(-1.81) (-0.62) (0.73) (0.67) (2.01) (-1.15) (2.41) (0.67)
Total Returns (screen +
weight + benchmark)
-1.05 0.72 0.95 1.32 0.81 -0.25 4.02 -0.62
(-0.80) (0.62) (0.88) (1.19) (0.51) (-0.26) (2.45) (-0.89)
Top - bottom
Screening
-2.36 -0.37 0.67 -1.82 -1.74 -3.03 -1.15 3.88
(-0.92) (-0.17) (0.51) (-1.02) (-0.65) (-1.42) (-0.66) (1.47)
Weighting
4.48 -1.18 0.51 0.32 -1.30 0.65 -1.85 1.23
(1.86) (-1.43) (0.80) (0.61) (-0.60) (0.86) (-1.35) (1.28)
Total Returns (screen +
weight + benchmark)
2.11 -1.55 1.18 -1.49 -3.04 -2.38 -3.00 5.11
(0.73) (-0.76) (0.95) (-0.83) (-1.06) (-1.13) (-1.34) (1.82)

97

Table 34. Average yearly gross returns (%) of funds grouped by their benchmark and market cap weighted screening intensity
The table displays the top and bottom decile portfolio returns, sorted based on the market cap weighted screening intensity for each category of funds. Fund returns
in each portfolio are weighted by their total assets. The bottom panel shows the results for the top-bottom portfolio. Finally, t-statistics are reported in brackets.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Top decile screening intensity
Screening
-1.67 -0.72 -0.61 1.45 0.02 -0.11 -6.24 -2.51 -1.93
(-1.77) (-0.58) (-0.41) (1.12) (0.02) (-0.08) (-2.42) (-0.92) (-0.81)
Weighting
0.57 0.88 1.05 0.33 -0.03 0.19 1.64 0.71 0.39
(1.21) (1.57) (1.21) (0.51) (-0.03) (0.26) (1.96) (1.07) (0.63)
Total Returns (screen +
weight + benchmark)
13.24 12.13 15.11 14.84 13.00 12.21 5.04 7.90 8.64
(4.26) (4.02) (4.47) (4.42) (3.49) (3.02) (0.81) (1.38) (1.49)
Bottom decile screening intensity
Screening
-0.36 -0.02 -0.36 0.29 0.94 -0.04 0.41 0.28 -0.37
(-1.03) (-0.08) (-0.42) (0.38) (1.29) (-0.05) (0.91) (0.28) (-0.46)
Weighting
0.55 0.25 0.95 0.68 0.14 -0.15 0.22 -0.07 -0.37
(1.55) (1.00) (1.68) (0.97) (0.43) (-0.45) (0.48) (-0.10) (-0.44)
Total Returns (screen +
weight + benchmark)
12.76 12.19 15.27 14.03 14.08 11.94 10.27 9.91 9.44
(4.92) (4.80) (4.49) (4.81) (4.03) (3.11) (2.09) (2.16) (2.04)
Top - bottom
Screening
-1.30 -0.70 -0.26 1.16 -0.91 -0.07 -6.66 -2.78 -1.56
(-1.54) (-0.57) (-0.19) (0.87) (-0.75) (-0.05) (-2.55) (-0.97) (-0.58)
Weighting
0.02 0.64 0.10 -0.35 -0.17 0.34 1.43 0.78 0.76
(0.07) (1.10) (0.13) (-0.53) (-0.20) (0.46) (1.82) (1.04) (1.11)
Total Returns (screen +
weight + benchmark)
0.48 -0.06 -0.16 0.81 -1.08 0.27 -5.23 -2.00 -0.80
(0.40) (-0.05) (-0.14) (0.70) (-0.97) (0.24) (-1.94) (-0.68) (-0.31)

98

Table 34 (continued). Average yearly gross returns (%) of funds grouped by their benchmark and market cap weighted
screening intensity
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Top decile screening intensity
Screening
-0.47 0.97 1.15 -1.35 -2.19 -2.52 1.00 4.97
(-0.26) (0.43) (0.73) (-0.67) (-0.67) (-1.13) (0.34) (1.55)
Weighting
1.46 0.52 -0.07 0.48 2.81 1.89 2.44 1.06
(1.03) (0.68) (-0.14) (0.76) (0.93) (2.56) (1.24) (0.88)
Total Returns (screen +
weight + benchmark)
7.62 12.52 12.40 8.84 3.38 9.69 12.31 17.42
(1.09) (2.21) (2.24) (1.57) (0.23) (1.98) (1.48) (3.26)
Bottom decile screening intensity
Screening
0.86 1.06 -0.30 -0.55 -1.48 -1.12 2.35 0.41
(0.72) (1.36) (-0.30) (-0.72) (-0.75) (-0.48) (1.17) (0.46)
Weighting
-0.55 -0.93 0.30 0.67 1.64 -0.06 -0.97 1.13
(-0.71) (-1.72) (0.61) (1.72) (1.43) (-0.14) (-1.80) (1.95)
Total Returns (screen +
weight + benchmark)
6.94 11.16 11.32 9.83 2.91 9.14 10.25 12.94
(0.95) (1.86) (1.92) (1.82) (0.22) (1.38) (1.25) (2.85)
Top - bottom
Screening
-1.33 -0.09 1.45 -0.80 -0.70 -1.41 -1.35 4.56
(-0.63) (-0.04) (1.13) (-0.39) (-0.17) (-0.37) (-0.38) (1.41)
Weighting
2.01 1.45 -0.37 -0.19 1.17 1.96 3.42 -0.07
(1.07) (1.77) (-0.65) (-0.30) (0.34) (2.20) (1.61) (-0.06)
Total Returns (screen +
weight + benchmark)
0.68 1.36 1.08 -0.99 0.47 0.55 2.07 4.48
(0.37) (0.65) (0.91) (-0.50) (0.09) (0.14) (0.66) (1.35)

99

Table 35. DGTW adjusted average yearly gross returns (%) of funds grouped by their benchmark and market cap weighted
screening intensity
The table displays the top and bottom decile portfolio returns, sorted based on the market cap weighted screening intensity for each category of funds. Fund returns
in each portfolio are weighted by their total assets. The bottom panel shows the results for the top-bottom portfolio. Finally, t-statistics are reported in brackets.
Overall S&P 500
S&P 500
Growth
S&P 500
Value
S&P
MidCap 400
S&P SmallCap
600
Russell 1000
Russell 1000
Value
Russell 1000
Growth
Top decile screening intensity
Screening
-0.66 -0.16 0.13 1.40 0.59 0.49 -4.24 -2.15 -1.41
(-1.14) (-0.21) (0.15) (1.67) (0.70) (0.64) (-2.13) (-1.04) (-0.77)
Weighting
0.18 0.61 0.19 -0.04 -0.18 -0.36 0.97 0.43 0.03
(0.76) (1.65) (0.44) (-0.10) (-0.41) (-0.83) (1.64) (0.80) (0.08)
Total Returns (screen +
weight + benchmark)
0.63 0.76 1.51 1.08 1.44 0.63 -2.29 -1.20 -0.34
(1.33) (1.05) (2.57) (1.34) (1.80) (0.94) (-1.25) (-0.61) (-0.21)
Bottom decile screening intensity
Screening
-0.20 0.01 -0.79 0.79 0.43 -0.15 0.77 0.44 0.08
(-0.87) (0.05) (-1.34) (1.42) (0.80) (-0.28) (1.63) (0.76) (0.14)
Weighting
0.16 0.09 0.49 -0.05 0.03 -0.51 -0.30 -0.27 -0.45
(0.87) (0.57) (1.30) (-0.12) (0.09) (-1.71) (-0.75) (-0.62) (-0.84)
Total Returns (screen +
weight + benchmark)
0.40 0.41 0.89 0.46 1.48 -0.15 1.45 0.68 0.66
(1.45) (1.36) (1.58) (0.67) (2.04) (-0.15) (2.41) (0.82) (0.64)
Top - bottom
Screening
-0.46 -0.17 0.92 0.61 0.17 0.64 -5.01 -2.58 -1.49
(-0.82) (-0.22) (1.14) (0.74) (0.23) (0.71) (-2.48) (-1.23) (-0.76)
Weighting
0.02 0.53 -0.30 0.01 -0.21 0.14 1.27 0.70 0.48
(0.07) (1.39) (-0.58) (0.03) (-0.42) (0.34) (1.82) (1.14) (0.83)
Total Returns (screen +
weight + benchmark)
0.23 0.36 0.61 0.62 -0.04 0.78 -3.74 -1.88 -1.01
(0.53) (0.49) (0.96) (0.86) (-0.05) (0.93) (-1.82) (-0.85) (-0.50)

100

Table 35 (continued). DGTW adjusted average yearly gross returns (%) of funds grouped by their benchmark and market cap
weighted screening intensity
Russell 2000
Russell 2000
Value
Russell 2000
Growth
Russell MidCap
Growth
Russell
MidCap
Russell
MidCap Value
Russell 3000
Wilshire
5000
Top decile screening intensity
Screening
0.97 1.88 1.06 -0.31 -1.85 -1.69 2.40 4.50
(0.72) (1.21) (1.00) (-0.20) (-0.98) (-1.05) (1.26) (1.77)
Weighting
0.45 0.15 -0.08 0.16 2.43 1.11 1.92 0.40
(0.50) (0.23) (-0.21) (0.41) (0.89) (1.89) (1.46) (0.43)
Total Returns (screen +
weight + benchmark)
0.29 2.46 2.22 -0.09 0.51 -0.22 3.39 4.85
(0.20) (1.60) (1.83) (-0.06) (0.13) (-0.12) (1.74) (1.91)
Bottom decile screening intensity
Screening
0.80 0.98 -0.13 0.01 0.31 -0.78 2.96 0.48
(0.73) (1.44) (-0.17) (0.01) (0.15) (-0.37) (1.59) (0.60)
Weighting
-0.91 -0.95 0.19 0.40 2.48 -0.52 -1.10 0.34
(-1.19) (-2.01) (0.44) (1.26) (1.40) (-1.21) (-2.28) (0.90)
Total Returns (screen +
weight + benchmark)
-1.25 0.47 1.30 0.47 2.72 -0.93 0.93 0.77
(-0.92) (0.39) (1.12) (0.42) (1.21) (-0.39) (0.56) (0.96)
Top - bottom
Screening
0.17 0.90 1.19 -0.31 -2.16 -0.91 -0.56 4.02
(0.12) (0.56) (1.35) (-0.19) (-0.69) (-0.30) (-0.22) (1.53)
Weighting
1.37 1.10 -0.27 -0.24 -0.05 1.63 3.02 0.06
(0.98) (1.45) (-0.56) (-0.48) (-0.01) (2.13) (1.95) (0.06)
Total Returns (screen +
weight + benchmark)
1.54 1.99 0.92 -0.55 -2.21 0.72 2.46 4.08
(0.95) (1.25) (0.97) (-0.33) (-0.46) (0.23) (0.94) (1.49)

101

XI. References
Amihud, Y. (2002). Illiquidity and stock returns: cross-section and time-series effects. Journal of
Financial Markets, 5, 31-56.
Baker, M., Litov, L., Wachter, J., Wurgler, J. (2010). Can mutual fund managers pick stocks?
Evidence from their trades prior to earnings announcements. Journal of Financial and
Quantitative Analysis, 45, 1111-1131.
Barnett, M., and Salomon, M. (2006). The curvilinear relationship between social responsibility
and financial performance. Strategic Management Journal, 27(11), 1101-1122.
Basu, S. (1977). Investment Performance of Common Stocks in Relation to their Price-Earnings
Ratio: A Test of the Efficient Market Hypothesis. Journal of Finance, 32, 663-682.
Becker, C., Ferson, W., Myers, D., and Schill, M. (1999). Conditional market timing with
benchmark investors. Journal of Financial Economics, 52, 119-148.
Capelle-Blancard, G., Monjon, S. (2012). The performance of socially responsible funds: does the
screening process matter? European Financial Management, 20(3), 494–520.
Carhart, M. (1997). On persistence in mutual fund performance. Journal of Finance, 52, 57-82.
Cohen, R. and Polk, C. (1998). The Impact of Industry Factors in Asset-Pricing Tests. Working
Paper. https://ssrn.com/abstract=7483
Cohen, R. and Polk, C. and Silli, B. (2010). Best Ideas. Working Paper.
http://dx.doi.org/10.2139/ssrn.1364827
Cremers, K. and Petajisto, A. (2009). How active is your fund manager? A new measure that
predicts performance. Review of Financial Studies, 22, 3329–65.
Daniel, K., Grinblatt, M., Titman, S., and Wermers, R. (1997). Measuring mutual fund
performance with characteristic-based benchmarks. Journal of Finance, 52, 1035–58.
102

Doshi, H., Elkamhi, R., and Simutin, M. (2015). Managerial activeness and mutual fund
performance. Review of Asset Pricing Studies, 5, 156-184.
Fama, E. and French, K. (1993). The Cross-Section of Expected Stock Returns. Journal of
Finance, 47, 427-465.
Fama, E. and French, K. (1996). Multifactor Explanations of Asset Pricing Anomalies. Journal of
Finance, 51, 55-84.
Fama, E. and French, K. (1997). Industry costs of equity. Journal of Financial Economics, 43,
153-193.
Ferson, W., and Khang, K. (2002). Conditional performance measurement using portfolio weights:
evidence for pension funds. Journal of Financial Economics, 65, 249-282.
Ferson, W., and Mo, H. (2016). Performance measurement with selectivity, market and volatility
timing. Journal of Financial Economics, 121, 93-110.
Grinblatt, M., and Titman, S. (1989). Mutual fund performance: an analysis of quarterly portfolio
holdings. Journal of Business, 62, 393-416.
Grinblatt, M., and Titman, S. (1993). Performance measurement without benchmarks: an
examination of mutual fund returns. Journal of Business, 66, 47-68.
Hjalmarsson, E. (2010). Predicting Global Stock Returns. Journal of Financial and Quantitative
Analysis, 45, 49-80.
Humphrey, J. and Lee, D. (2011). Australian socially responsible funds: performance, risk and
screening intensity. Journal of Business Ethics, 102(4), 519-535.
Jensen, M. (1968). The performance of mutual funds in the period 1945-1964. Journal of Finance,
23, 389-416.
Jiang, H., Verbeek, M. and Wang, Y. (2014). Information content when mutual funds deviate from
benchmarks. Management Science, 60, 2038-2053. doi:
http://dx.doi.org/10.1287/mnsc.2013.1847
103

Kacperczyk, M., Van Nieuwerburgh, S. and Veldkamp, L. (2016). A Rational Theory of Mutual
Funds' Attention Allocation. Econometrica, 84, 571-626.
Nieuwerburgh, S., Veldkamp, L. (2010). Information Acquisition and Under-
Diversification. Review of Economic Studies, 70(2), 779-805.
Sensoy, B. (2009). Performance evaluation and self-designated benchmark indexes in the mutual
fund industry. Journal of Financial Economics, 92, 25–39.
Shawky, H. and Smith, D. (2005). Optimal Number of Stock Holdings in Mutual Fund Portfolios
Based on Market Performance. The Financial Review, 40, 481-495.
Stambaugh, R. (1999). Predictive Regressions. Journal of Financial Economics, 54, 375-421.
Wermers, R. (2004). Is Money Really Smart? New Evidence On The Relation Between Mutual
Fund Flows, Manager Behavior, and Performance Persistence. Working Paper.
http://dx.doi.org/10.2139/ssrn.414420
104

Appendix A: An example of a fund’s prospectus
Figure A1. An example of a fund’s prospectus

105

Figure A1 (continued). An example of a fund’s prospectus

Abstract (if available)

Abstract This paper develops a holdings-based measure of fund performance that distinguishes how fund managers weight stocks in their portfolios from how they screen the stocks they choose to hold. I find that screening decisions contribute negatively to the performance of a typical fund whereas portfolio weighting decisions contribute positively. In particular, screening decisions lower fund performance by 0.40% per year before costs. Weighting decisions contribute 0.72% in performance per year for a typical fund during 1980-2016. Even though the managers possess weighting ability, which in isolation suggests skill, when I also consider the ability to pick which stocks to hold, skill is no longer present. My results also suggest that fund managers could improve their performance by following a benchmark closer in terms of holdings but not in terms of weights.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Essays in tail risks

PDF

Essays on financial markets

PDF

Two essays on financial econometrics

PDF

Two essays on the mutual fund industry and an application of the optimal risk allocation model in the real estate market

PDF

Essays in asset pricing

PDF

Evolution of returns to scale and investor flows during the life cycle of active asset management

PDF

Essays in empirical asset pricing

PDF

Expectation dynamics and stock returns

PDF

Investment behavior of mutual fund managers

PDF

Workplace organization and asset pricing

PDF

Essays on delegated portfolio management under market imperfections

PDF

Three essays in derivatives, trading and liquidity

PDF

Essays in financial intermediation

PDF

Empirical analysis of factors driving stock options grants and firms volatility

PDF

Examining managerial and interpersonal skills delivery in MBA education: an evaluation study

PDF

Disclosure distance and earnings announcement returns

PDF

How do acquirers govern the deal-making process? Three essays on U.S. mergers and acquisitions 1994 – 2017

PDF

How do securities analysts respond to different types of news events?

PDF

Innovation: financial and economics considerations

PDF

The term structure of CAPM alphas and betas

Asset Metadata

Creator Skripnik, Roman (author)

Core Title Mutual fund screening versus weighting

School Marshall School of Business

Degree Doctor of Philosophy

Degree Program Business Administration

Publication Date 06/18/2019

Defense Date 05/13/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag active management,managerial skill,OAI-PMH Harvest,screening,weighting

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Ferson, Wayne (committee chair), Jones, Christopher (committee member), Sloan, Richard (committee member), Zapatero, Fernando (committee member)

Creator Email eng2rom@yahoo.com,skripnik@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-176422

Unique identifier UC11660814

Identifier etd-SkripnikRo-7503.pdf (filename),usctheses-c89-176422 (legacy record id)

Legacy Identifier etd-SkripnikRo-7503.pdf

Dmrecord 176422

Document Type Dissertation

Format application/pdf (imt)

Rights Skripnik, Roman

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA