1. Investment Viewpoints
Portfolio Turnover:
It’s Not As Important As You May Think
By Kevin Lewis, CFA, Vice President, Client Portfolio Manager
FOR INSTITUTIONAL USE ONLY/NOT FOR PUBLIC USE
2. Kevin is Client Portfolio Manager for the Growth Equity Investment
Discipline at American Century Investments. Prior to his current role,
he was a portfolio manager for two domestic growth strategies at
American Century Investment Management. Before joining the firm
in 1995, Kevin was a senior equity portfolio manager at Virtus Capital
Management and has worked in the investment industry since 1983.
Kevin earned a bachelor’s degree in business from Virginia Tech. He is a
CFA charterholder and a member of the CFA Institute.
Kevin Lewis, CFA
Vice President
Client Portfolio Manager
FOR INSTITUTIONAL USE ONLY/NOT FOR PUBLIC USE 2
3. Several recent academic research papers have rejected the widely-held notion that portfolio
turnover has a negative impact on investment performance. Some of the reasons given for this
contradictory assertion include the facts that:
• portfolio turnover is not a good proxy for trading costs
• trading costs have declined over time due to improvements in trading technology combined with
the rise of alternative trading markets
• there are deficiencies in the calculation of the “portfolio turnover” statistic itself.
Their final conclusion, and one that we were able to corroborate with our own independent
brief study, is that advisors and consultants should utilize other statistics to evaluate investment
managers because higher portfolio turnover is not correlated with lower investment returns.
The portfolio turnover and performance debate
High portfolio turnover has long been considered a drag on investment returns. Academic studies
dating back fifteen years provided evidence that this relationship once existed. In those studies,
higher turnover had a negative correlation with returns. In particular, an article in a 1993 issue of
The Review of Financial Studies by E. Elton, M. Gruber, S. Das and M. Hlavka notes that historical
performance was negatively correlated with portfolio turnover.
But more recent academic studies have arrived at different—and conflicting—conclusions. Between
2000 and 2001, four new academic studies on the subject were published. Two (Russ Wermers in
a 2000 Journal of Finance article, and John Chalmers, Roger Edelen and Gregory Kadlec in a 2001
University of Pennsylvania Wharton Working Paper) concluded that there was no broad statistical
relationship between turnover and investment returns. They contended that the turnover and return
relationship varied considerably depending on the market capitalization and investment style of
portfolios being considered. For example, large-cap growth portfolios had substantially different
correlations of turnover and returns than those for small-cap value portfolios.
Meanwhile, two other papers concluded that portfolio turnover is positively correlated with
investment performance. The idea behind these papers was to try to distinguish between turnover
based on information content, that is, “informed turnover,” seeking opportunities to generate
alpha, versus “uninformed turnover,” based on other motivations such as investment flows, risk
management requirements and ex-post reactions to sudden changes in security values.
Most recently, a March 2007 paper titled Scale Effects in Mutual Fund Performance: The Role of
Trading Costs1 was posted for peer review, and generated substantial interest both in academic and
industry circles. The paper described an eleven-year analysis of mutual fund returns and confirmed
that portfolio turnover is not generally harmful to fund returns. This conclusion rested on the
observation that the components of trading costs vary substantially based on the size and style of the
portfolio and trade motives.
FOR INSTITUTIONAL USE ONLY/NOT FOR PUBLIC USE 3
4. The 2007 Study
The March 2007 paper describes a comprehensive study that directly estimated the trading costs
of 1,706 open-ended mutual funds of all size/style categories from 1995 to 2005. Morningstar
supported the research by providing access to its databases, allowing researchers to track quarterly
trades for every security held by each mutual fund in the study universe over an eleven-year
timeframe. Dr. Gregory Kadlec, one of the authors and professor of finance at Virginia Tech,
presented the paper at the June 2007 Morningstar Investors Conference.
The authors’ motive for this study was the growing recognition that portfolio turnover does not
necessarily result in higher trading costs. To back this assertion, they cite the conflicting conclusions
of past studies and make the following points:
• Turnover is a poor proxy for determining actual fund trading costs because unit trading costs
vary dramatically across size/style categories of securities. Referencing a 2004 industry study that
estimated per-unit trading costs for large-cap value stocks at 32 basis points (bps) and for small-
cap growth stocks at 132 bps, they observe that a large-cap value portfolio with 150% turnover will
have lower annual trading costs than a small-cap growth portfolio with 50% turnover.
• Because turnover is reported as a percentage of average assets instead of absolute dollar volume,
it fails to capture differences in trading costs arising from differences in trade size. The authors
reference both theoretical and empirical studies that find trade size is an important determinant of
trading costs.
• Turnover does not capture all of a portfolio’s trading activity. Defined as the minimum of security
purchase or sales divided by average fund assets for a given period, turnover statistics can be
misleading. For example, a portfolio with purchases of 100% of average assets and sales of 20%
will have lower reported turnover than another portfolio with purchases and sales both equal to
25% of average assets.
Given these limitations of turnover, the authors note that it’s not surprising that the conclusions of
past studies on turnover and returns have been mixed.
They defined an analytical model to bypass turnover and directly estimate portfolio trading costs:
Trading Costs = Trading Volume x Per Unit Trading Costs
where trading volume equals the size of each trade, and per unit trading costs reflect commissions
paid, the bid-ask spread of the trade and the price impact of each trade. The former two factors—
commissions and the bid-ask spread— reflect prior insights that the ‘cost of turnover’ varies
substantially based on the style/style of shares traded. The latter factor— the price impact of
trading— reflects the prior insight that trade motive has a significant impact on trading cost. The
paper concludes that all of these factors in combination define a statistical relationship with portfolio
returns that turnover fails to capture.
FOR INSTITUTIONAL USE ONLY/NOT FOR PUBLIC USE 4
5. Portfolio turnover impact in the recent environments
Given the market environment in the latter part of 2007 and first half of 2008, we wanted to see if
the newer academic research findings persisted, particularly since this period was noted for its
substantial market volatility and uncertainty. Our team reviewed the reported mutual fund turn-
over and returns based on the Morningstar size/style matrix for the twelve month period ending
June 30th, 2008. The analysis included 2,529 open-ended U.S. mutual funds and compared turnover
versus returns for each of the nine size/style categories. Our objective was to determine if we
could detect a statistical relationship between the turnover and returns in this volatile environment.
(See Appendix for more detail on the methodology used in our study).
Extreme values— however unusual— can skew regression analysis. So we also ran our regression
excluding the highest decile of turnover in each of the nine size/style categories. The graphs below
illustrate the regression results for large-cap growth portfolios for the period from July 1, 2007 to
June 30, 2008. Notice in both graphs that the regression line slope is slightly positive, indicating
weak positive relationship between turnover and subsequent performance.
Large Cap Growth: 1 Year Returns vs. Portfolio Turnover Large Cap Growth: 1 Year Returns vs. Portfolio Turnover
(All Deciles of Fund Turnover Included) (Highest Decile of Fund Turnover Excluded)
July 1, 2007 to June 30, 2008 July 1, 2007 to June 30, 2008
Large Cap Growth: 1 Yr Fund Returns vs. Turnover Large Cap Growth: 1 Yr Fund Returns vs. Turnover
All Funds Included Highest Decile of Fund Turnover Excluded
30.0 30.0
20.0 20.0
10.0 10.0
1 Year Fund Return
1 Year Fund Return
0.0 0.0
0 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 0 20 40 60 80 100 120 140 160 180 200
-10.0 -10.0
1 Year Fund Returns 1 Year Fund Returns
-20.0 -20.0
-30.0 -30.0
-40.0 -40.0
FundPortfolio Annual Turnover(%)
Annual Turnover (%) Fund AnnualTurnover (%) (%)
Portfolio Annual Turnover
Source: Morningstar and American Century Investments
Similar regressions were run for the other eight size/style fund categories. The size/style matrix
on the next page summarizes our review. We found that there was no relationship between the two
variables in the majority of the categories. The large cap growth category showed a weak positive
statistical correlation between turnover and returns. And the mid cap growth catagory showed a
weak negative statistical correlation which reversed to a weak positive correlation when funds in the
top decile of turnover were eliminated. Therefore the data from this more-volatile-than-average one-
year period did not appear to deviate from longer-term academic research findings. That is, portfolio
turnover in the recent environment continued to be a poor predictor of portfolio trading costs and
investment performance.
FOR INSTITUTIONAL USE ONLY/NOT FOR PUBLIC USE 5
6. Study Summary: Turnover vs. Returns
Value Core Growth
No relationship between portfolio No relationship between portfolio Weak positive relationship
Large Cap
turnover and performance turnover and performance between portfolio turnover and
returns.
o relationship between portfolio
N o relationship between portfolio
N Weak negative relationship*
turnover and performance turnover and performance (all deciles of turnover) or weak
Mid Cap
positive relationship (top deile
of turnover excluded) between
turnover and returns
o relationship between portfolio o relationship between portfolio No relationship between portfolio
Small Cap
N N
turnover and performance turnover and performance turnover and performance
* Due to Outlier Issues in Highest Turnover Decile of Funds
Source: Morningstar and American Century Investments
Implications for investment professionals
In assessing portfolio manager performance, turnover is a poor indicator of actual fund trading costs
that can hurt investment returns. Recent academic research, as well as our own study, reject the long-
standing belief that higher fund turnover usually results in lower returns.
Using turnover as a proxy for trading costs in evaluating asset managers is like using height alone to
judge the skill and athleticism of basketball players. Financial professionals shouldn’t be automati-
cally biased toward “low turnover mangers”. Instead, they should judge whether the manager’s
investment process improves the odds for excess returns.
Bear in mind that discretionary trading motivated by information and portfolio manager skill has
lower trading costs than large asset flow-driven trades. Investors should be able to count on a
manager to make beneficial trades based on his experience and current information.
And remember that a firm’s trading desk can be an important contributor to portfolio performance
(and a competitive advantage) through smart execution that minimizes the cost components of a
trade. The past decade has seen vast improvements in trading technology as new alternative trading
markets (e.g. ECNs) have been created. These have helped lower trading costs for many managers.
In summary, portfolio turnover has been shown to be unrelated with performance. Using it to
screen for asset managers could lead to the elimination of those with the potential to produce excess
returns. Therefore, consultants and advisors should use other screening criteria to judge a manager’s
investment process and ability to produce excess returns.
Please contact your American Century Investments representative with questions or comments.
FOR INSTITUTIONAL USE ONLY/NOT FOR PUBLIC USE 6
7. Appendix
I: Past Research on the Effects of Turnover on Fund Performance
Study Authors/Journal Year Conclusions
Elton, Gruber, Das and Hlavka 1993 Fund returns are negatively related to turnover
Journal of Financial Studies
Carhart 1997 Fund returns are negatively correlated to turnover
Journal of Finance
Dahlquist, Engstrom and Soderlind 2000 Fund returns are positively related to turnover
Journal of Financial and Quantitative Analysis
Wermers 2000 No relationship between fund returns and turnover
Journal of Finance
Chalmers, Edelen, Kadlec 2001 No relationship between fund returns and turnover
UPenn Wharton White Paper
Chen, Jagadeesh and Wermers 2001 Fund returns are positively related to turnover
Journal of Financial and Quantitative Analysis
Table Source: Evans, Edelen and Kadlec March 2007 paper
II: Our Study Methodology for Comparing Turnover and Fund Performance
We undertook a statistical analysis of reported mutual fund turnover and returns based on the
Morningstar size/style matrix using one year data from July 1st, 2007 to June 30th, 2008. The
analysis included 2,529 open-ended U.S. mutual funds. We regressed fund turnover versus returns
for each of the nine size/style categories to determine if we could detect a relationship.
The fund turnover data for our study universe ranged from under 5% per year to over 4,000%.
Extreme values— however unusual— can skew regression analysis in a phenomenon called the
‘outlier effect’. So we elected to run our regressions twice: Once with the full range of funds and
turnover, and a second time excluding funds that comprised the highest decile of turnover in each of
the nine size/style categories. The figures below illustrate the regression results for mid cap growth
funds in both cases:
Mid Cap Growth: 1 Year Returns vs. Portfolio Turnover Mid Cap Growth: 1 Year Returns vs. Portfolio Turnover
(All Deciles of Fund Turnover Included) (Highest Decile of Fund Turnover Excluded)
July 1, 2007 to June 30, 2008
Large Cap Growth: 1 Year Fund Returns vs. Turnover
July 1, 2007 to June 30, 2008
Large Cap Growth: 1 Year Fund Returns vs. Turnover
(All Deciles of Fund Turnover Included) (Highest Decile of Fund Turnover Excluded)
30.0
Mid Cap Growth: 1 YrMarch 31,Returns vs. Turnover
April 1, 2007 to
Fund 2008 30.0 Mid Cap Growth: 2007 to MarchReturns vs. Turnover
April 1, 1 Yr Fund 31, 2008
All Funds Included Highest Decile of Fund Turnover Excluded
20.0 20.0
20.0 20.0
10.0 10.0
10.0 10.0
0.0 0.0
0 200 400 600 800 1,000 1,200 1,400 0 20 40 60 80 100 120 140 160 180 200
1 Year Fund Return
0.0 0.0
1 Year Fund Return
-10.0 0 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 -10.00 50 100 150 200
-10.0 -10.0
-20.0 -20.0
1 Year YearReturns (%)
1 Fund Fund Returns
1Year Fund Returns (%)
1 Year Fund Returns
-20.0 -20.0
-30.0 -30.0
-30.0 -30.0
-40.0 -40.0
-40.0 -40.0
-50.0 -50.0
Fund Annual Turnover (%) Fund Annual Turnover (%)
Fund Annual Turnover(%)
Portfolio Annual Turnover (%) Portfolio Annual Turnover (%)
Fund Annual Turnover (%)
FOR INSTITUTIONAL USE ONLY/NOT FOR PUBLIC USE 7
8. In the left hand chart, a total of 282 mid cap growth funds covering the full range of reported
turnover were included. Note the outlier effect mentioned earlier due to one fund with nearly
1,350% annual turnover—DireXion Spectrum Equity Opportunity (SFEOX). In the right hand chart,
funds within the highest decile of turnover (29 total) were excluded, bringing the turnover range
down to more realistic values with a maximum of 200% per year.
This kind of paired regression was run for all of the other eight size/style fund categories. Our basic
statistical test involved a null hypothesis that stated “There is no significant relationship
between fund turnover and returns”. Using statistical tests of significance, we calculated the ‘risk’ of
being wrong if we rejected this null hypothesis and concluded instead that a relationship did exist.
Statisticians call this alpha risk and use a metric called the ‘p-Value’ to quantify it.
A very low p-Value (under 5%) is a common threshold used to justify rejecting the null hypothesis.
And in those cases where our analysis led us to do so (concluding a relationship between fund
turnover and returns did exist) we calculated two other statistics: The slope of the regression line
to determine the direction and sensitivity of the relationship between turnover and returns; and the
R-squared (R2) which defines how much of the total variation in fund returns is explained by the
variation in turnover.
The size/style matrix below summarizes our findings. It was only with mid cap growth funds in-
cluding the top decile of turnover that we could statistically establish a weak relationship for
the ‘accepted wisdom’ that higher fund turnover results in lower returns. However the strength
of this negative relationship reversed and became a positive relationship with the highest decile
excluded (29 funds with average turnover of 381% per year).
Study Summary Statistics by Size/Style Category
Value Core Growth
All Deciles x Top Decile
E All Deciles Ex Top Decile All Deciles Ex Top Decile
p-Value 62.0% 53.0% 62.0% 55.0% 1.9% 8.0%
Large
Cap
R 2
0.4% 0.0% 1.0% 0.6% 0.1% 1.6%
Slope 0.8 -0.2 0.5 1.2 0.1 2.0
p-Value 7.4% 3.0% 18.0% 16.0% 1.6% 1.0%
Mid
Cap
R 2
0.1% 0.0% 0.1% 0.0% 0.2% 4.9%
Slope 0.3 0.0 0.1 0.0 -0.2 3.5
p-Value 17.1% 26.0% 1.7% 27.0% 94.0% 90.0%
Small
R2 1.9% 0.4% 2.1% 3.3% 0.0% 0.1%
Cap
Slope -0.8 -1.4 -0.9 -3.5 0.1 0.7
Note: Slope is expressed as basis points of return for each additional 1% increase in fund turnover
A final finding was that, in those cases where we could statistically establish a relationship
between fund turnover and returns (positive or negative), the R2 of the relationship (or the
explanatory power of variations in turnover to account for variations in returns) was very low.
FOR INSTITUTIONAL USE ONLY/NOT FOR PUBLIC USE 8