Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment - - PowerPoint PPT Presentation

▶
fee structure return chasing and mutual fund choice an
SMART_READER_LITE
LIVE PREVIEW

Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment - - PowerPoint PPT Presentation

Apr 23, 2023 10 likes •351 views

Research Question Experimental Design Experimental Results Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment Mikhail Anufriev a Te Bao b Angela Sutan c Jan Tuinstra d a University of Technology Sydney b University of Groningen

slide-1
SLIDE 1

Research Question Experimental Design Experimental Results

Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

Mikhail Anufrieva Te Baob Angela Sutanc Jan Tuinstrad

aUniversity of Technology Sydney bUniversity of Groningen cBurgundy School of Business dUniversity of Amsterdam

ESA, Sydney July 24, 2015

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-2
SLIDE 2

Research Question Experimental Design Experimental Results

Plan

Research Question Experimental Design Experimental Results

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-3
SLIDE 3

Research Question Experimental Design Experimental Results

Research Topic: Mutual Fund Choice

◮ $30 trillions of assets were invested in mutual funds ◮ 46, 3% of US households own assets ◮ median value $100, 000

(data from Investment Company Institute for 2013 year)

Q: What are the main factors that drive people’s choice decision?

  • 1. role of fee
  • 2. role of past returns

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-4
SLIDE 4

Research Question Experimental Design Experimental Results

Answers from recent empirical literature

◮ Barber, Odean and Zheng (2005): flow of money is negatively

correlated with the front-end load but not correlated with

  • peration expenses fee

◮ people avoid paying salient and transparent fee

◮ Khorana and Servaes (2012): front-end load fee funds

dominate over operation fee funds

◮ front-end load fee is a commitment device to reduce search

cost

◮ Carhart (2007), Gil-Bazo and Ruiz-Verd´

u (2009): fees charged by the fee is negatively correlated with returns

◮ Sirri and Tufano (1998), Choi, Laibson, Madrian and Metrick

(2009): past performance of funds and past personal experience affect peoples’ decisions

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-5
SLIDE 5

Research Question Experimental Design Experimental Results

Answers from recent experimental literature

◮ Wilcox (2003): too much attention is paid to past

performance and to the front-load fee, too little to operational expenses

◮ one period decision, cross-sectional comparison of choices

◮ Choi, Laibson and Madrian (2010): people rely on (irrelevant)

information about past realized returns and invest in the high past performance/high fee funds

◮ one period decision, between-treatments comparison

◮ Ehm and Weber (2013, survey): a higher propensity to choose

fund with a performance fee (fraction of gains) over

  • perational expenses fee

◮ loss aversion

◮ Bloomfield and Hales (2002): prediction of random walk

realization

◮ employ various prediction methods on the basis of patters Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-6
SLIDE 6

Research Question Experimental Design Experimental Results

This paper

◮ sequential decision making on binary choice ◮ known Data Generation Processes for funds A and B ◮ 3 treatments: No fee, Operational expenses, Front-end load. ◮ the same optimal choice (invest in B) in all treatments

Main Results:

  • 1. Both the fee structure and uninformative past returns affect

decisions

  • 2. No “fee aversion”: subjects are attracted to the fund that has

higher gross return but charges a fee.

  • 3. No “salient fee aversion”: front-end load fee and operational

fee attract comparable number of subjects in average

  • 4. Front-end load leads to “locking” many (but not all) subjects

in the fund with the fee

  • 5. Reinforcement learning from past returns

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-7
SLIDE 7

Research Question Experimental Design Experimental Results

Plan

Research Question Experimental Design Experimental Results

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-8
SLIDE 8

Research Question Experimental Design Experimental Results

General Settings

◮ Two mutual funds: A and B. ◮ Subjects know the DGP of returns of each fund. E.g., in N:

◮ Net return of A is 0.03 + ǫA,t. Net return of B is 0.04 + ǫB,t. ◮ ǫA,t and ǫB,t are i.i.d. shocks that take 0.02 and −0.02 with

equal probability.

◮ The rational choice is to choose B always.

◮ Subjects have initial wealth, M0 = 1, 000 points ◮ Decision: Every period all accumulated wealth can be invested

in A or in B or in neither of them.

◮ Subject’s wealth grows by the net return rate of the chosen

fund (doesn’t change if neither of the funds has been chosen).

◮ Design: 3 treatments, 3 blocks of DGP with 15 periods each

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-9
SLIDE 9

Research Question Experimental Design Experimental Results

Idea behind Treatment Design

◮ Let PX,t be the price of mutual fund X in period t. ◮ DGP assumes the gross return

RX,t = PX,t PX,t−1 =

  • 1 + gX + ε

with probability 1

2

1 + gX − ε with probability 1

2 ◮ the τ-periods expected return is

Et

  • R(τ)

X,t

  • = (1 + gX)τ .

◮ In treatment O, the operational fee γXPX,t−1 is paid:

Et

  • R(τ)

O,X,t

  • = (1 + gX − γX)τ

◮ In treatment F, the front-end load fee FXMt should be paid.

Et

  • R(τ)

F,X,t

  • = (1 − FX) Et
  • R(τ)

X,t

  • = (1 − FX) (1 + gX)τ

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-10
SLIDE 10

Research Question Experimental Design Experimental Results

Idea behind Treatment Design

RX,t = PX,t PX,t−1 = 1 + gX + ε

w.p. 1

2

1 + gX − ε w.p. 1

2

, Et

  • R(τ)

X,t

  • = (1 + gX)τ

Et

  • R(τ)

O,X,t

  • = (1 + gX − γX)τ ,

Et

  • R(τ)

F,X,t

  • = (1 − FX) (1 + gX)τ

◮ Treatment N: None of the two funds, A and B, require fees.

We chose ε = 0.02, gA = 0.03 and gB = 0.04 so that ε < gA < gB and ε > 1

2(gB − gA) ◮ Treatment O: Fund A is as before. Fund B gives higher

return, gB = 0.05, but requires operating expenses γB = 0.01.

◮ Treatment F: Fund A is as before. Fund B gives higher

return, gB = 0.05, but requires front-end load fee of 0.13.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-11
SLIDE 11

Research Question Experimental Design Experimental Results

Overview of experiment

Fund A Fund B Trtmnt gA E1

  • R(14)

A,1

  • − 1

gB γB FB E1

  • R(14)

B,1

  • − 1

N 3% 51.26% 4% − − 73.17% O 3% 51.26% 5% 1% − 73.17% F 3% 51.26% 5% − 13% 72.25% Choice B is optimal at every period in N, O and F. Switch to B is optimal in F until period 7. Participants: 1st year Master students, Burgundy School of Business with 22/19/35 participants in treatments.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-12
SLIDE 12

Research Question Experimental Design Experimental Results

Price and Return Time Series in the first 15 periods

treatment N treatments O and F Returns of investing for 15 periods in A is 45.1%; in B is 86.2%

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-13
SLIDE 13

Research Question Experimental Design Experimental Results

Price and Return Time Series in the middle 15 periods

treatment N treatments O and F Returns of investing for 15 periods in A is 50.9%; in B is 53.9%

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-14
SLIDE 14

Research Question Experimental Design Experimental Results

Price and Return Time Series in the last 15 periods

treatment N treatments O and F Returns of investing for 15 periods in A is 34.3%; in B is 66.2%

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-15
SLIDE 15

Research Question Experimental Design Experimental Results

Experimental Screen

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-16
SLIDE 16

Research Question Experimental Design Experimental Results

Testable Hypotheses

◮ N and O differ only in framing.

Hypothesis 1. There is no significant difference in subjects’ behavior between treatments N and O.

◮ B is rational choice in all treatments; and there is (almost) no

difference in expected return of choosing B between O and F. Hypothesis 2. There is no significant difference in subjects’ behavior between treatments F and O.

◮ return realizations are IID

Hypothesis 3. For treatments N and O the fraction of choices for fund B in any particular period is uncorrelated with the realized returns of funds A and B in the previous period.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-17
SLIDE 17

Research Question Experimental Design Experimental Results

Plan

Research Question Experimental Design Experimental Results

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-18
SLIDE 18

Research Question Experimental Design Experimental Results

Treatment Periods Fraction of Choosing A B Neither N Period 1 45.45% 54.55% 0.00% Block 1 41.52% 58.48% 0.00% Block 2 48.79% 50.91% 0.30% Block 3 27.27% 72.73% 0.00% Average 39.19% 60.71% 0.10% O Period 1 68.42% 31.58% 0.00% Block 1 32.63% 67.02% 0.35% Block 2 50.88% 48.77% 0.35% Block 3 20.35% 79.65% 0.00% Average 34.62% 65.15% 0.23% F Period 1 37.14% 62.86% 0.00% Block 1 24.38% 74.48% 1.14% Block 2 37.90% 60.76% 1.33% Block 3 24.76% 74.86% 0.38% Average 29.01% 70.03% 0.95% Populational Fractions

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-19
SLIDE 19

Research Question Experimental Design Experimental Results

What do the Numbers Say

◮ Some participants choose neither (0.1% to 1.33%), even if

choosing A generates strictly positive return.

◮ Many participants choose A with high frequency (20% to

50%), even if choosing B generates a higher expected return.

◮ Fraction of choice of A is the lowest in block 3, though it

increases from block 1 to block 2.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-20
SLIDE 20

Research Question Experimental Design Experimental Results

Comparing N and O

Treatment Periods Fraction of Choosing A B Neither N Block 1 41.52% 58.48% 0.00% Block 2 48.79% 50.91% 0.30% Block 3 27.27% 72.73% 0.00% Average 39.19% 60.71% 0.10% O Block 1 32.63% 67.02% 0.35% Block 2 50.88% 48.77% 0.35% Block 3 20.35% 79.65% 0.00% Average 34.62% 65.15% 0.23% Result 1. We reject Hypothesis 1 and find that in treatment O there is a significantly higher fraction of choices for fund B than in treatment N. (at 5% level according to the Wilcoxon test)

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-21
SLIDE 21

Research Question Experimental Design Experimental Results

Comparing O and F

Treatment Periods Fraction of Choosing A B Neither O Block 1 32.63% 67.02% 0.35% Block 2 50.88% 48.77% 0.35% Block 3 20.35% 79.65% 0.00% Average 34.62% 65.15% 0.23% F Block 1 24.38% 74.48% 1.14% Block 2 37.90% 60.76% 1.33% Block 3 24.76% 74.86% 0.38% Average 29.01% 70.03% 0.95% Saliency of the fee has a limited influence on aggregate investment behavior.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-22
SLIDE 22

Research Question Experimental Design Experimental Results

Heterogeneity in Individual Choice

There is a lot heterogeneity in individual choice in F.

◮ most subjects choose B in the first or second period and stick

with this choice until the end

◮ some subjects choose B for a limited number of periods

Histogram of individual choice of B. Result 2. We reject Hypothesis 2: there is a significant difference, at the individual level, of subjects’ behavior in treatments O and F.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-23
SLIDE 23

Research Question Experimental Design Experimental Results

Average number of switches

Treatment Periods N O F Block 1 0.28 0.20 0.22 Block 2 0.32 0.30 0.18 Block 3 0.27 0.20 0.12 All Blocks 0.29 0.23 0.17 Frequency of switching per subject per period.

◮ Differences between treatments are significant at 5% level. ◮ The number of switches in F is the smallest. ◮ Some subjects switch in F more than twice, i.e., paid fee

twice.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-24
SLIDE 24

Research Question Experimental Design Experimental Results

Number of switches (over all blocks)

Empirical CDF of total number of switches by each individual.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-25
SLIDE 25

Research Question Experimental Design Experimental Results

Number of times subjects paid the front-end load fee

Occasions of choosing B after investing in A T B 1 2 3 4 5 F All 6 (6%) 73 (70%) 14 (13%) 7 (7%) 4 (4%) 0 (0%) F 1 22 4 5 3 2 3 22 8 1 1 3 3 29 2 1 N All 0 (0%) 12 (18%) 17 (25%) 24 (36%) 10 (15%) 3 (5%) O All 0 (0%) 21 (37%) 18 (32%) 10 (18%) 7 (12%) 1 (2%) Treatment F is substantially different from the other two:

◮ out of 73 single investments in B, 59% lasted for the full block and

73% lasted for at least 12 periods

◮ The front-end load locks a lot of subjects into the choice of B ◮ There is a learning effect

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-26
SLIDE 26

Research Question Experimental Design Experimental Results

Earnings

The empirical CDF of individual earnings (in terms of points).

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-27
SLIDE 27

Research Question Experimental Design Experimental Results

Earnings and Efficiency

Treatment N O F F1 F2 Mean Block 1 1653.86 1682.81 1565.55 1781.37 1200.33 (75.8%) (79.2%) (65.6%) (90.6%) (23.2%) Block 2 1503.22 1493.52 1385.09 1464.53 1186.46 (93.4%) (91.6%) (71.5%) (86.2%) (34.6%) Block 3 1593.08 1579.89 1553.51 1584.74 1220.33 (89.6%) (87.6%) (83.6%) (88.3%) (33.3%) All Blocks 4750.17 4756.22 4504.15 4804.35 3591.93 (84.8%) (85.1%) (73.0%) (87.5%) (28.7%) Average individual earnings in different treatments and blocks. (Efficiency is defined as a ratio of earnings towards always investing in B)

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-28
SLIDE 28

Research Question Experimental Design Experimental Results

Return Chasing

The time series of actual fraction choice of B against the binary variable that indicates whether the realized return of B is larger than A in the last period.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-29
SLIDE 29

Research Question Experimental Design Experimental Results

Return Chasing

Treatment N O F Block 1 59.09% (41.23%) 26.32% (30.08%) 32.86% (23.88%) Block 2 61.36% (47.73%) 57.90% (50.00%) 41.43% (39.39%) Block 3 25.33% (25.76%) 12.28% (16.92%) 26.67% (24.29%) All Blocks 51.24% (38.10%) 39.71% (32.33%) 35.84% (29.18%) The fraction of choices for fund A in periods immediately following a period in which fund A had a higher realized return than fund B. The fraction in between brackets refer to the (unconditional) fraction of choices for fund A in that block.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-30
SLIDE 30

Research Question Experimental Design Experimental Results

Discrete Choice Model

nB,t = exp [β0 + β1 (rB,t−1 − rA,t−1)] 1 + exp [β0 + β1 (rB,t−1 − rA,t−1)] . Treatment N O F β0 0.3798 0.5689 0.7956 p-value (0.0000) (0.0000) (0.0000) β1 0.1423 0.0975 0.0628 p-value (0.0000) (0.0000) (0.0000) McFadden’s R2 0.0327 0.0145 0.0070 Estimated coefficients of the logit model. Result 3. We reject Hypothesis 3. Fund choices are, to a substantial extent, explained by past realized returns.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-31
SLIDE 31

Research Question Experimental Design Experimental Results

Fitted Time Series in N

The actual and fitted fraction of B.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-32
SLIDE 32

Research Question Experimental Design Experimental Results

Fitted Time Series in O

The actual and fitted fraction of B.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-33
SLIDE 33

Research Question Experimental Design Experimental Results

Fitted Time Series in F

The actual and fitted fraction of B.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment

slide-34
SLIDE 34

Research Question Experimental Design Experimental Results

Conclusion

◮ Making rational fund choice decision is a difficult task. ◮ O vs. N: Subjects pay attention to gross expected return,

ignoring operational fee.

◮ F vs. O: Front-end load fee is not more discouraging than

  • peration fee but may play a role of commitment device

locking subject in fund B.

◮ We find evidence for return chasing and described the data

with a simple discrete choice model.

Anufriev, Bao, Sutan & Tuinstra UTS, UoG, UvA & ESC-Dijon Fee Structure, Return Chasing, and Mutual Fund Choice: an Experiment