Internet Appendix for:

The Worst, the Best, Ignoring All the Rest: The Rank Effect and Trading Behavior This Draft: August 21st, 2014 This Appendix presents additional results and robustness analysis to accompany the paper “The Worst, The Best, Ignoring All the Rest: The Rank Effect and Trading Behavior.” Figure IA.1 This figure graphs the probability of sale by rank for individual investors by rank allowing for a different impact of rank for different portfolio sizes. Thus all of the graphs represent coefficients from the same regression of sell on rank interacted with portfolio size, stock by day by holding period fixed effects and additional controls. Figure IA.2 This figure graphs the probability of sale by rank for individual investors using the regression from Table 4 column [3]. Table IA.1 This table repeats the analysis examining the probability of sale with controls for performance since purchase from Table 4 using returns with trading fees subtracted. Table IA.2 This table examines the probability of sale with controls for performance since purchase from Table 4, but uses alternative rankings. Columns [1] through [4] rank position based on the overall magnitude of a gain or loss. Columns [5] through [8] rank positions based on the average daily return over the period a position has been held. Table IA.3 This table repeats the analysis examining the probability of sale with controls for performance since purchase from Table 4 columns [3] adding controls to proxy for changes in the expectation of a stock’s mean and. To control for changes in means, a proxy for recent news is included which is the cumulative abnormal return (CAR) from two days before the most recent earnings announcement to 1 day after (similar to Frazzini 2006). Dummy variables for being in the highest and lowest two deciles (based on the previous month) are included to control for particularly good and bad news that likely resulted in shifts in expected future returns. In order to analyze shifts in variance, changes of the variance from a recent period are compared to a prior period. A short term measure of the variance is constructed using the variance over the previous month (the preceding 1-20 trading days) divided by the variance in the month before that (the preceding 21-40 trading days). A long term measure examines the prior 6 months (the preceding 1-125 trading days) divided by the variance in the 6 months prior to that (the preceding 126-250 trading days).

Table IA.4 This table repeats the analysis examining the probability of sale with controls for performance since purchase from Table 4 columns [3] adding controls to proxy for changes to a stock’s covariance with economic state variables. I utilize the state variables specified in Petkova (2006), the market return (RM), dividend yield of the market (Div), term spread (Term), default spread (Def), risk free rate (RF) and return on the HML (RHML) and SMB portfolios (RSMB). These variables are modeled using a VAR framework as follows: 𝑹𝑴,𝒕 𝑹𝑴,𝒕−𝟏 ⎡ ⎤ ⎡ ⎤ ⎢ 𝑫𝒊𝒗𝒕 ⎥ ⎢ 𝑫𝒊𝒗𝒕−𝟏 ⎥ ⎢𝑻𝒆𝒓𝒎𝒕 ⎥ ⎢𝑻𝒆𝒓𝒎𝒕−𝟏 ⎥ ⎢ 𝑫𝒆𝒇𝒕 ⎥ = 𝑨 ⎢ 𝑫𝒆𝒇𝒕−𝟏 ⎥ + 𝒖𝒕 ⎢ 𝑹𝑭 ⎥ ⎢ 𝑹𝑭 ⎥ 𝒕 𝒕−𝟏 ⎢ ⎥ ⎢ ⎥ ⎢ 𝑹𝑯𝑴𝑳,𝒕 ⎥ ⎢ 𝑹𝑯𝑴𝑳,𝒕−𝟏 ⎥ ⎣ 𝑹𝑺𝑴𝑩,𝒕 ⎦ ⎣ 𝑹𝑺𝑴𝑩,𝒕−𝟏 ⎦

Innovations in these variables, the “surprise series” from Petkova, are multiplied with the stock return to obtain estimates of changes to the covariance of the stock with respect to the state variables. Petkova utilizes monthly data, but this study uses daily. To mimic the monthly specification, the data is used in 20 day trading increments, starting the day before a sell day. Thus the calculation is analogous to using monthly data, but avoids any issues of staleness as it is always calculated the day before the sell day. Table IA.5 This table reports analysis similar to Table 4, but utilizing a cox proportional hazard model rather than a logit specification. The specification allows for a different strata for each account (WFICN). Table IA.6 This table reports analysis similar to Table 4, but utilizing nearest neighbor propensity score matching rather than a logit specification. The propensity score analysis is run as follows. Logit regressions are run separately for each day. A dummy for best-ranked or worst-ranked is regressed on Return, √Holding Days*Return, Variance and √Holding. Positions are matched with replacement to the next closest propensity score that is not extreme-ranked on that day. Table IA.7 This table controls for covariate balance in the extreme-ranked and non-extreme-ranked groups utilizing entropy balancing (Hainmueller 2012). The balancing is conducted on return, square root of holding days, variance and the interaction of the return and holding days. The balancing is conducted separately for each day which indirectly controls for changes over time of the effect of the covariates. To make sure there is enough data for balancing, in the investor data I exclude days in the bottom quartile of observations per day (626 observations).

Table IA.8 This table repeats the analysis examining the probability of sale with controls for performance since purchase from Table 4 using all of the variables on the left hand side from Ben-David and Hirshleifer (2012). The additional variables include the log of the purchase price, separate dummies for zero returns, and interactions of Gain, Loss, and Zero with holding period. Table IA.9 Ben-David and Hirshleifer (2012) find that the propensity to sell stocks is non-linear in returns. This table examines the probability of sale with controls for performance since purchase from Table 4 column [3] and [6] allowing for quadratic, cubic and quartic patterns in returns to control for such non-linearity. Table IA.10 This table examines the probability of sale with controls for performance since purchase from Table 4 column [3] for portfolios of different sizes. Column [1] limits the sample to portfolios with 3 to 4 stocks using the specification of Table 4 column [2]. Column [2], [3], and [4] use the specification of Table 4 column [3] restricting the sample to portfolios containing 5 to 8 stocks, 9 to 13 stocks and 14 or more stocks respectively. Table IA.11 This table examines the probability of sale with controls for performance since purchase from Table 4 column [3], splitting the sample by the time since the investor last traded (time between a sell day and the previous day a position was bought or sold). Column [1] examines 1 to 20 days, column [2] examines 21 to 250 days, column [2] examines 251 or more days, and column [4] includes the days since previous trade variable as a control. Table IA.12 This table repeats the analysis examining the probability of sale with controls for performance since purchase from Table 4 with a dummy for liquidation on the left hand side rather than sell. Table IA.13 The rank effect may not capture a propensity to sell extremes, but rather an increased propensity to sell a stock throughout the entire portfolio as it moves closer to best or worst rank. This table examines the probability of sale with controls for performance since purchase from Table 4 column [3] adding controls for the percentile of rank in the portfolio. Column [1] allows for different effects for a position at a gain or at a loss, while column [2] allows for different effect depending on whether a position is above, or below the median portfolio rank. The table suggests it is extreme ranks, not percentiles that accounts for the rank effect. Table IA.14 This table repeats the analysis examining the probability of sale with controls for performance since purchase from Table 4 allowing for a different impact of rank based on length of time the position is held.

Table IA.15 This table repeats the analysis examining the probability of sale with controls for performance since purchase from Table 4 adding a third cluster group of stocks. Thus the standard errors are triple clustered by account (wficn), date and stock. Table IA.16 This table repeats the analysis examining the probability of sale with controls for performance since purchase from Table 4 expanding the sample to include all days a position can be sold, regardless of whether it is a sell day. This is similar to how Ben-David and Hirshleifer (2012) examine the data. Table IA.17 This table repeats the analysis examining the probability of sale with controls for performance since purchase from Table 4 columns [3] and [6] separately for each year in the sample. Table IA.18 This table repeats the analysis examining the probability of sale with controls for performance since purchase from Table 4 columns [3] separately by sex and by quartile of age. Table IA.19 This table repeats the analysis examining the probability of sale with controls for performance since purchase from Table 4 columns [3] separately by self-reported income and net worth. Table IA.20 This table examines the probability of sale with controls for performance since purchase from Table 4 column [3], but examines performance in excess of the market return. Column [1] adds market return since purchase as a control. Column [2] uses return in excess of the market return as a control and calculates gain and loss based on this variable. Column [3] uses these controls and ranks the stocks in the portfolio based on this return as well. Table IA.21 This table repeats the analysis for individual investors examining the probability of sale with controls for performance since purchase from Table 4 column [3] including a variety of fixed effects. Column [1] includes account fixed effects, column [2] includes stock fixed effects, column [3] includes account and stock fixed effects, column [4] includes account by day fixed effects, column [5] includes stock by day fixed effects and column [6] includes account by day and stock by day fixed effect. Table IA.22 This table repeats the for mutual funds analysis examining the probability of sale with controls for performance since purchase from Table 4 column [6] including a variety of fixed effects. Column [1] includes fund fixed effects, column [2] includes stock fixed effects, column [3] includes fund and stock fixed effects, column [4] includes fund by day fixed

effects, column [5] includes stock by day fixed effects and column [6] includes fund by day and stock by day fixed effect. Table IA.23 This table repeats the analysis examining the probability of sale with controls for performance since purchase from Table 4 columns [3] adding controls for the realized capital gain or loss within a tax year. Selling might occur to mitigate the overall size of a taxable gain or loss in a tax year. If an individual realizes a net capital gain during a tax year they can realize losses to offset this gain and minimize taxes. Similarly, if an investor realizes a net capital loss in a tax year, they can sell positions at a gain with less of a tax consequence. The benefits of realizing positions can be different at above or below $3,000, so differences in effect above and below this threshold are controlled for. Table IA.24 This table repeats the analysis examining alphabetical ordering of Table 8 ordering by ticker instead of company name. Table IA.25 This table analyzes the decision to purchase more of a position in a portfolio. It presents summary statistics and repeats the analysis examining the probability of sale with controls for performance since purchase from Table 4 using buy instead of sell. Table IA.26 This table repeats the analysis examining the impact of various aspects of extremeness on trading behavior from Table 9. Measures scaled by standard deviation in Table 9 are measured in levels in this table. Measures in levels in Table 9 are scaled by the standard deviation in this table. Table IA.27 This table contains the analysis examining various aspects of extremeness on trading behavior from Table 9, for buying instead of selling. Table IA.28 This table repeats analysis similar to Table 4, but examines the fraction of a position sold or the decision to liquidate by mutual funds as the dependent variable. Table IA.29 This table repeats analysis similar to Table 9, but on mutual funds rather than individual investors. Table IA.30 This table examines the optimality of the rank effect by examining performance in excess of the market after positions are sold. Similar to previous studies, momentum is the dominant factor. Thus, worst-ranked positions typically perform underperform the market over the next year after they are sold. Best-ranked positions typically outperform the market over the next year after they are sold. Panel A examines excess returns of sold positions that are best worst and middle ranked. Panel B examines the difference in returns between sold best and worst positions and the portfolio of everything else that is not sold.

References Ben-David, Itzhak and David Hirshleifer. 2012, “Beyond the Disposition Effect: Do Investors Realy Like Realizing Gains More Than Losses,” Review of Financial Studies, 25(8), 2485-2532. Frazzini, Andrea, 2006, “The Disposition Effect and Under-Reaction to News”, Journal of Finance, 61, 2017-2046. Hainmueller, Jens, 2012, Entropy Balancing for Causal Effects: A Multivariate Reweighting Method to Produce Balanced Samples in Observational Studies, Political Analysis 20(1) 25-46. Petokova, Ralitsa, 2006, “Do the Fama-French Factors Proxy for Innovations in Predictive Variables,” Journal of Finance, 62(2), 581-612.

Figure IA.1 – The Rank Effect for Individual Investors by Portfolio Size 5 to 6 stocks

4th Best

3rd Best

2nd Best

Best

4th Best

2nd Best

Best

2nd Best

Best

9 to 10 stocks

3rd Best

Worst

4th Best

Best

3rd Best

4th Worst

3rd Worst

2nd Worst

0 Worst

Best

3rd Best

Worst

2nd Best

3rd Worst

Worst

2nd Worst

0

.05

.05

.1

.1

.15

.15

.2

.2

.25

7 to 8 stocks

Best

.15 .1 .05

Best

Worst

13 to 14 stocks

5th Best

6th Best

6th Worst

5th Worst

4th Worst

3rd Worst

2nd Worst

Worst

Best

3rd Best

2nd Best

0

Worst

4th Best

5th Best

5th Worst

4th Worst

3rd Worst

Worst

2nd Worst

0

.05

.1

.15

11 to 12 stocks

Best

Best

Worst

5th Best

6th Best

7th Best

7th Worst

6th Worst

5th Worst

4th Worst

3rd Worst

2nd Worst

Worst

Best

3rd Best

2nd Best

4th Best

5th Best

6th Best

0

Worst

7th Best

7th Worst

6th Worst

5th Worst

4th Worst

3rd Worst

Worst

2nd Worst

0

.02

.02

.04

.04

.06

.06

.08

.08

.1

.1

15 or more stocks

Best

This table presents results from a linear probability model. The dependent variable is a dummy variable equal to 1 if a stock is sold and is regressed on rank variables interacted with the number of stocks in the portfolio. Each dummy variable for rank is interacted with the indicated number of stocks in the chart title. Fixed effects for the interaction of CUSIP, date and holding period decile are included. Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

Figure IA.2 – Seasonality of the Rank Effect for Individual Investors

Marginal effect from logit regression as specified in Table 4 column [3]. Regressions are run separately for each calendar month. Best is the coefficient on the dummy variable for the highest return and worst is the coefficient from the dummy variable for the lowest ranked return. The dotted line is the upper and lower bound of the 95% confidence interval. This figure contains individual trading data from January 1991 through November 1996.

Table IA.1 – The Rank Effect for Investors Incorporating Fees into Returns [1] Best Worst

[2] 0.156 (25.48) 0.110 (15.02)

2nd Best 2nd Worst Return*Gain Return*Loss Gain Return*Gain *√Holding Days Return*Loss *√Holding Days Variance *Gain Variance *Loss √Holding Days Observations R2

0.033 (3.37) -0.113 (-4.95) 0.047 (13.55) -0.001 (-4.07) 0.002 (1.59) 7.412 (2.93) -2.632 (-2.21) -0.002 (-9.44) 1,048,549 0.012

-0.015 (-1.87) -0.002 (-0.09) 0.035 (10.96) -0.001 (-2.32) 0.001 (0.83) 5.641 (2.64) -2.592 (-2.64) -0.002 (-10.16) 1,048,549 0.034

[3] 0.204 (26.10) 0.150 (15.79) 0.121 (19.84) 0.088 (12.11) -0.030 (-4.00) 0.036 (1.77) 0.030 (9.94) 0.000 (-1.41) 0.000 (0.18) 4.863 (2.56) -2.480 (-2.98) -0.002 (-10.97) 1,048,549 0.049

This table presents marginal effects from logit regressions. The dependent variable is a dummy variable equal to 1 if a stock is sold. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase. Stocks is the number of positions held in a portfolio on a sell day. Data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

Table IA.2 – Ranking by Alternative Variables

Best Worst

Size of Gain/Loss Individual Investor Mutual Fund

Individual Investor

[1] 0.175 (25.03) 0.104 (21.28)

[5] 0.136 (21.87) 0.075 (16.52)

2nd Best 2nd Worst Return*Gain Return*Loss Gain Additional Controls Observations R2

0.000 (0.02) -0.054 (-2.83) 0.029 (8.14) X 1,048,549 0.038

[2] 0.223 (25.30) 0.141 (21.08) 0.127 (17.59) 0.085 (14.58) -0.013 (-1.78) -0.023 (-1.26) 0.025 (7.67) X 1,048,549 0.053

[3] 0.218 (26.16) 0.121 (9.66)

0.024 (4.82) -0.257 (-11.89) -0.013 (-3.89) X 15,603,277

[4] 0.224 (26.45) 0.124 (9.67) 0.171 (21.58) 0.104 (10.04) 0.018 (3.60) -0.247 (-11.49) -0.013 (-3.85) X 15,603,277

0.007

0.008

Average Daily return

Mutual Fund

0.017 (2.20) -0.097 (-4.73) 0.026 (7.33) X 15,603,394

[6] 0.180 (20.88) 0.109 (16.99) 0.108 (14.44) 0.079 (13.31) 0.007 (0.97) -0.074 (-3.77) 0.022 (6.48) X 15,603,394

[7] 0.059 (9.53) 0.070 (7.14)

0.031 (6.34) -0.256 (-11.42) -0.013 (-4.07) X 15,603,394

[8] 0.061 (9.54) 0.073 (7.14) 0.047 (7.89) 0.055 (6.18) 0.029 (5.97) -0.246 (-10.81) -0.013 (-4.10) X 15,603,394

0.027

0.039

0.006

0.006

This table presents marginal effects from logit regressions. The dependent variable is a dummy variable equal to 1 if a stock is sold. In columns 1-4 Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) dollar gain/loss in the portfolio. In columns 5-8 Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) average daily return in the portfolio. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase. Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Investor data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. Mutual fund data are from report dates from January 1990 to June 2010. A fund must hold at least 20 CRSP merged securities to be included in the analysis. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account for the investors and date and WFICN for the mutual fund data.

Table IA.3 – Rank Effect with Controls for Mean, Variance and Past Performance Best Worst 2nd Best 2nd Worst High CAR Low CAR

[1] 0.203 (22.41) 0.147 (20.12) (0.13) (17.99) (0.08) (14.75) 0.010 (6.83) 0.007 (5.14)

Short Term: High ΔVar

Long Term: High ΔVar Long Term: Low ΔVar

R2

[3] 0.202 (20.73) 0.145 (19.70) 0.124 (16.32) 0.084 (14.13)

0.014 (10.90) -0.003 (-3.18)

Short Term: Low ΔVar

Other Controls Observations

[2] 0.203 (20.89) 0.146 (19.83) 0.125 (16.45) 0.084 (14.21)

X 949,484 0.046

0.008 (6.60) -0.001 (-0.70) X X 1,034,179 1,017,467 0.047

0.046

[4] 0.202 (22.27) 0.146 (19.94) 0.126 (17.85) 0.084 (14.62) 0.010 (6.35) 0.007 (4.65) 0.013 (10.08) -0.004 (-3.73) 0.008 (6.66) 0.001 (0.86) X 939,226 0.047

This table presents the marginal effects from logit regressions of a dummy variable equal to 1 if a stock is sold on characteristics of the stock being held. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. Stocks are not included on the day their position is opened. Standard errors are clustered by account and date. Best (Worst) is a dummy variable equal to 1 if the stock has the best (worst) return in the portfolio. Best (Worst) Ret is the highest (lowest) return in a portfolio and 2nd Best (Worst) Ret is the second highest (lowest). For the cumulative abnormal returns from the most recent earnings announcement (CAR) and the variance (Var) variables, high refers to the highest quintile and low refers to the lowest. CAR is calculated from two days before the most recent earnings announcement to one day after. Short term Var is the Var calculated over the previous month divided by the variance from two months prior. Long term is the previous 6 months divided by the 6 months prior. Other Controls include Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Data covers January 1991 to November 1996.

Table IA.4 – Rank Effect with Controls for Covariance and Past Performance

Best Worst 2nd Best 2nd Worst

Mkt Div Term Default Rf High hml High smb Other Controls Observations R2

Level

Short Term

Long Term

[1] 0.201 (20.97) 0.147 (20.12) 0.123 (16.48) 0.085 (14.43) High Low 0.009 0.001 (6.28) (0.80) 0.002 0.004 (1.45) (2.31) 0.004 0.002 (2.88) (1.71) 0.005 0.003 (3.77) (2.11) 0.004 0.008 (2.98) (5.25) 0.006 0.009 (4.14) (5.38) 0.008 0.001 (3.82) (0.50) X 1,048,514

[2] 0.203 (21.20) 0.148 (20.14) 0.124 (16.58) 0.085 (14.41) High Low 0.005 0.005 (3.62) (3.27) 0.002 0.003 (1.74) (2.13) 0.004 0.004 (3.47) (3.27) 0.002 0.001 (1.57) (1.01) 0.003 0.004 (2.06) (3.06) 0.003 0.003 (2.51) (2.88) 0.002 0.003 (1.36) (2.20) X 1,014,555

[3] 0.201 (20.90) 0.146 (19.71) 0.123 (16.38) 0.084 (14.23) High Low 0.003 0.001 (2.34) (0.59) 0.000 0.002 (0.36) (1.49) 0.001 0.002 (0.47) (1.01) 0.004 0.005 (2.93) (3.36) 0.005 0.002 (3.54) (1.83) 0.006 0.005 (3.07) (4.08) 0.002 0.002 (1.45) (1.55) X 980,126

0.049

0.047

0.047

This table presents the marginal effects from logit regressions of a dummy variable equal to 1 if a stock is sold on characteristics of the stock being held. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included. Stocks are not included on the day their position is opened. Standard errors are clustered by account. Best (Worst) is a dummy variable equal to 1 if the stock has the best (worst) return in the portfolio. Best (Worst) Ret is the highest (lowest) return in a portfolio and 2nd Best (Worst) Ret is the second highest (lowest). See the appendix description at the beginning of this document for the details of the how the covariance variables are calculated. Level is the most recent value, short term is the ratio of the most recent value and one month prior while long term is the ratio of the most recent value and the value one year earlier. High is a dummy for the top quintile and low is a dummy for the lowest quintile. Other Controls include Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Data covers January 1991 to November 1996.

12

Table IA.5 – Hazard Model Individual Investor [1] Best Worst

[2] 0.501 (20.08) 0.486 (19.68)

2nd Best 2nd Worst Return*Gain Return*Loss Gain Variance *Gain Variance *Loss Observations R2

-0.842 (-21.11) 1.704 (22.83) 0.128 (3.75) 52.155 (3.16) 2.853 (1.24) 1,048,549 0.013

-1.089 (-22.93) 2.340 (26.72) 0.099 (2.90) 50.034 (3.04) 4.050 (1.93) 1,048,549 0.017

Mutual Fund

[3] 0.672 (22.98) 0.639 (21.85) 0.364 (19.15) 0.303 (16.11) -1.180 (-22.99) 2.503 (27.41) 0.087 (2.59) 49.344 (3.01) 4.384 (2.14) 1,048,549 0.018

[4]

[5] 0.058 (3.71) 0.180 (17.79)

-0.233 (-23.45) 0.149 (8.74) -0.158 (-30.93) 47.032 (14.50) 3.435 (5.67) 15,603,394

-0.238 (-21.40) 0.184 (10.19) -0.160 (-31.08) 46.977 (14.46) 3.386 (5.84) 15,603,394

[6] 0.057 (3.36) 0.191 (17.99) -0.022 (-1.83) 0.138 (15.56) -0.237 (-20.16) 0.204 (10.87) -0.162 (-31.23) 46.952 (14.45) 3.394 (5.90) 15,603,394

0.002

0.002

0.002

This table presents the coefficients from a cox proportional hazard model with a separate strata for each account (WFICN). Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. Stocks are not included on the day that position is opened. A stock is considered best (worst) if the stock has the highest (lowest) return in the portfolio. Investor data covers January 1991 to November 1996. Mutual fund data are from January 1990 to June 2010. Dates examined are report dates. A fund must hold at least 20 CRSP merged securities to be included in the analysis. The top number is the coefficient, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by account and date.

13

Table IA.6 – Rank Effect using Propensity Score Matching Individual Investor

Best - Not Best Worst - Not Worst

Entropy Balanced 0.128 (21.36) 0.084 (18.85)

Propensity Score 0.122 (29.65) 0.082 (23.24)

Mutual Fund Entropy Balanced 0.116 (12.40) 0.165 (10.03)

Propensity Score 0.103 (24.62) 0.143 (25.63)

This table presents the proportion of best (worst) positions sold versus the proportion of not best (worst) positions sold. “Entropy Balanced” is the weighted average difference based on entropy balancing of the sample each day taken from Table 15. The propensity score column is the difference between the proportion (best) worst sold and a sample of nearest neighbor matches that are not best (worst) ranked. Propensity scores are from logit regressions, separately each day of Best (Worst) on Return, √Holding Days*Return, Variance and √Holding. Not best (worst) stocks are matched to the closest propensity score with replacement. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. Stocks are not included on the day that position is opened. A stock is considered best (worst) if the stock has the highest (lowest) return in the portfolio. Investor data covers January 1991 to November 1996. Mutual fund data are from January 1990 to June 2010. Dates examined are report dates. A fund must hold at least 20 CRSP merged securities to be included in the analysis. The top number is the difference, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by account (WFICN) and date.

14

Table IA.7 – Rank Effect after Entropy Balancing Individual Investor Unweighted Best - Not Best 0.120 (19.74) Worst - Not Worst 0.059 (9.15)

Entropy Balanced 0.128 (21.36) 0.084 (18.85)

Mutual Fund Unweighted 0.116 (15.12) 0.188 (20.74)

Entropy Balanced 0.116 (12.40) 0.165 (10.03)

This table presents the proportion of best (worst) positions sold minus the proportion of not best (worst) positions sold. The “Unweighted” column is the simple difference. “Entropy Balanced” is the weighted average difference based on entropy balancing of the sample each day. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. Stocks are not included on the day that position is opened. A stock is considered best (worst) if the stock has the highest (lowest) return in the portfolio. Investor data covers January 1991 to November 1996. Mutual fund data are from January 1990 to June 2010. Dates examined are report dates. A fund must hold at least 20 CRSP merged securities to be included in the analysis. The top number is the difference, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by account and date.

15

Table IA.8 – Rank Effect with Ben-David and Hirshleifer (2012) Controls Investor [1] Best Worst

[2] 0.158 (20.43) 0.107 (19.79)

2nd Best 2nd Worst Return*Gain Return*Loss Gain Zero Log(Buy Price) Return*Gain *√Holding Days Return*Loss *√Holding Days Variance *Gain Variance *Loss Zero*√Holding Days Gain* √Holding Days √Holding Days Observations R2

0.040 (4.08) -0.186 (-9.12) 0.054 (8.90) -0.031 (-5.06) -0.001 (-0.53) -0.001 (-4.35) 0.006 (6.81) 5.500 (2.84) -3.226 (-3.11) 0.002 (3.81) -0.001 (-5.10) -0.001 (-2.91) 1,048,549 0.010

-0.008 (-1.00) -0.072 (-3.75) 0.050 (8.92) -0.029 (-5.19) 0.002 (2.67) -0.001 (-2.22) 0.005 (6.76) 4.402 (2.82) -2.354 (-2.96) 0.002 (3.79) -0.002 (-6.71) -0.001 (-2.29) 1,048,549 0.032

Mutual Fund [3] 0.207 (21.46) 0.147 (19.86) 0.127 (16.79) 0.084 (14.15) -0.024 (-3.37) -0.033 (-1.79) 0.047 (8.98) -0.028 (-5.20) 0.004 (4.76) 0.000 (-1.05) 0.005 (6.15) 3.869 (2.76) -1.925 (-2.91) 0.002 (4.11) -0.002 (-7.25) -0.001 (-2.30) 1,048,549 0.047

[4]

[5] 0.106 (11.89) 0.159 (12.59)

0.054 (11.14) -0.277 (-10.58) -0.011 (-2.02) -0.024 (-1.67) 0.044 (9.22) -0.001 (-6.72) 0.008 (9.06) 11.956 (5.11) -0.042 (-0.09) 0.001 (0.61) 0.000 (-0.79) -0.002 (-5.42) 15,603,394

0.045 (8.85) -0.247 (-9.65) -0.012 (-2.30) -0.027 (-1.86) 0.044 (9.26) -0.001 (-5.70) 0.008 (8.99) 12.048 (5.17) 0.021 (0.05) 0.001 (0.62) 0.000 (-0.67) -0.002 (-5.42) 15,603,394

[6] 0.117 (12.28) 0.165 (12.68) 0.106 (13.01) 0.119 (10.70) 0.037 (7.31) -0.230 (-8.99) -0.012 (-2.31) -0.028 (-1.91) 0.043 (9.31) 0.000 (-4.95) 0.007 (8.95) 12.067 (5.22) 0.139 (0.33) 0.001 (0.62) 0.000 (-0.69) -0.002 (-5.42) 15,603,394

0.009

0.009

0.010

This table presents marginal effects from logit regressions. The dependent variable is a dummy variable equal to 1 if a stock is sold. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase. Zero is a dummy variable equal to one if a position has a return of exactly zero. Investor data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. Mutual fund data are from report dates from January 1990 to June 2010. A fund must hold at least 20 CRSP merged securities to be included in the analysis. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account for the investors and date and WFICN for the mutual fund data.

16

Table IA.9 – The Rank Effect with Non-Linear Return Controls Individual Investor Best Worst 2nd Best 2nd Worst Pos*Return Pos*Return2

[1] 0.209 (21.33) 0.144 (19.50) 0.128 (16.78) 0.083 (13.83) -0.026 (-3.74) 0.003 (7.52)

[2] 0.214 (21.37) 0.144 (19.36) 0.130 (16.97) 0.082 (13.62) -0.039 (-5.60) 0.008 (4.92) 0.000 (-2.73)

-0.053 (-2.31) -0.119 (-5.41)

-0.148 (-3.96) -0.469 (-4.94) -0.338 (-3.98)

0.032 (9.32) X 1,034,212

0.037 (10.24) X 1,034,212

[3] 0.217 (21.39) 0.144 (19.35) 0.131 (17.01) 0.081 (13.61) -0.050 (-6.82) 0.015 (9.19) -0.001 (-5.46) 0.000 (3.84) -0.214 (-4.14) -0.887 (-3.66) -1.233 (-2.59) -0.580 (-1.93) 0.041 (10.94) X 1,034,212

0.047

0.047

0.047

Pos*Return3 Pos*Return4 Neg*Return Neg*Return2 Neg*Return3 Neg*Return4 Gain Additional Controls Observations R2

Mutual Fund [4] 0.120 (11.83) 0.171 (12.28) 0.106 (12.50) 0.123 (10.42) 0.017 (3.13) 0.000 (0.17)

[5] 0.120 (11.73) 0.170 (12.16) 0.106 (12.39) 0.122 (10.37) 0.017 (3.14) 0.000 (0.20) 0.000 (-0.20)

[6] 0.119 (11.68) 0.169 (12.13) 0.105 (12.08) 0.122 (10.36) 0.019 (3.17) 0.000 (-1.23) 0.000 (1.64) 0.000 (-1.78) -0.256 -0.376 -0.292 (-8.77) (-12.17) (-7.32) -0.053 -0.438 0.037 (-1.88) (-4.63) (0.22) -0.314 0.572 (-3.71) (1.70) 0.512 (2.33) -0.010 -0.003 -0.007 (-3.52) (-1.10) (-2.46) X X X 15,603,394 15,603,394 15,603,394 0.007

0.007

0.007

This table presents marginal effects from logit regressions. The dependent variable is a dummy variable equal to 1 if a stock is sold. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase. Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Investor data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. Mutual fund data are from report dates from January 1990 to June 2010. A fund must hold at least 20 CRSP merged securities to be included in the analysis. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account for the investors and date and WFICN for the mutual fund data.

17

Table IA.10 – The Rank Effect for Individual Investors by Number of Stocks in the Portfolio

Best Worst

3-4 Stocks

5-8 Stocks

9-13 Stocks

14+ Stocks

[1] 0.146 (33.65) 0.023 (5.90)

0.122 (8.48) -0.110 (-3.52) 0.078 (16.17) X 201,873

[2] 0.159 (33.98) 0.079 (16.05) 0.070 (22.20) 0.023 (7.40) 0.038 (3.13) -0.096 (-3.68) 0.045 (11.80) X 317,953

[3] 0.104 (17.40) 0.095 (14.73) 0.065 (16.06) 0.039 (10.09) 0.029 (2.81) -0.033 (-1.23) 0.031 (8.13) X 233,138

[4] 0.076 (8.85) 0.094 (11.17) 0.067 (8.91) 0.052 (9.92) -0.003 (-0.39) -0.035 (-1.42) 0.010 (2.77) X 483,121

0.031

0.026

0.022

0.020

2nd Best 2nd Worst Return*Gain Return*Loss Gain Additional Controls Observations R2

This table presents marginal effects from logit regressions. The dependent variable is a dummy variable equal to 1 if a stock is sold. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase. Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

18

Table IA.11 – The Rank Effect Splitting by Days Since Last Trade

Best Worst 2nd Best 2nd Worst

1-20 Days

21-250 Days

251+ Days

Days on RHS

[1] 0.190 (17.97) 0.130 (17.90) 0.122 (15.39) 0.076 (13.42)

[2] 0.202 (41.22) 0.138 (26.79) 0.108 (30.47) 0.067 (18.82)

[3] 0.141 (8.50) 0.137 (6.97) 0.063 (4.69) 0.074 (5.09)

-0.030 (-3.76) -0.012 (-0.56) 0.027 (7.15) X 723,316

0.024 (2.92) -0.042 (-2.17) 0.015 (5.41) X 296,576

0.061 (1.60) -0.232 (-2.23) 0.005 (0.35) X 14,320

[4] 0.197 (22.85) 0.136 (21.01) 0.119 (17.89) 0.075 (14.78) 0.029 (32.51) -0.017 (-2.41) -0.001 (-0.07) 0.024 (7.77) X 1,034,212

0.055

0.038

0.017

0.053

Days since trade (100s) Return*Gain Return*Loss Gain Additional Controls Observations R2

This table presents marginal effects from logit regressions. The dependent variable is a dummy variable equal to 1 if a stock is sold. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase. Days since trade is the sell date minus the date of the last trade (either a buy or sell) divided by 100. Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

19

Table IA.12 – The Rank Effect for Investors Examining Liquidations Rather than Sells [1] Best Worst

[2] 0.139 (20.85) 0.091 (20.48)

2nd Best 2nd Worst Return*Gain Return*Loss Gain Return*Gain *√Holding Days Return*Loss *√Holding Days Variance *Gain Variance *Loss √Holding Days Observations R2

0.041 (4.43) -0.146 (-8.76) 0.032 (10.87) -0.002 (-6.20) 0.003 (3.65) 2.649 (2.61) -3.916 (-3.83) -0.002 (-11.68) 1,048,549 0.013

-0.009 (-1.21) -0.048 (-3.09) 0.027 (9.60) -0.001 (-4.52) 0.002 (2.90) 1.900 (2.06) -3.487 (-4.05) -0.002 (-12.11) 1,048,549 0.036

[3] 0.184 (21.95) 0.125 (20.66) 0.108 (17.62) 0.070 (15.08) -0.026 (-3.74) -0.016 (-1.11) 0.024 (9.32) -0.001 (-3.72) 0.001 (2.22) 1.489 (1.73) -3.204 (-4.18) -0.002 (-12.70) 1,048,549 0.051

This table presents marginal effects from logit regressions. The dependent variable is a dummy variable equal to 1 if a stock is liquidated. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase. Stocks is the number of positions held in a portfolio on a sell day. Data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

20

Table IA.13 – The Rank Effect Controlling For Rank Throughout the Entire Portfolio [1]

[2]

Best Worst 2nd Best 2nd Worst Gain*Rank Pctile Loss*Rank Pctile

-0.092 (-9.34) 0.115 (8.55)

(Pctile<.5)*Rank Pctile

0.015 (1.45) -0.057 (-2.54) 0.149 (11.68)

-0.047 (-3.95) -0.035 (-4.84) 0.034 (3.28) -0.172 (-7.90) 0.029 (7.61)

1,034,212 0.014

(Pctile>.5)*Rank Pctile Return*Gain Return*Loss Gain Additional Controls Observations R2

[3] 0.236 (17.70) 0.146 (14.81) 0.142 (15.97) 0.079 (11.95) 0.056 (6.87) -0.020 (-1.78)

[4] 0.261 (20.29) 0.117 (14.76) 0.142 (18.16) 0.072 (11.81)

1,034,212

-0.006 (-0.86) -0.016 (-0.82) -0.016 (-1.44) X 1,034,212

0.156 (14.64) 0.073 (9.63) -0.005 (-0.70) 0.010 (0.54) 0.027 (8.91) X 1,034,212

0.010

0.047

0.048

This table presents marginal effects from logit regressions. The dependent variable is a dummy variable equal to 1 if a stock is sold. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase. Percentile is the rank of a stock (where best is ranked 1) divided by the number of positions in the portfolio. Thus the percentile of best-ranked stocks is close to 0, while the percentile of worst-ranked stocks is close to 1. Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

21

Table IA.14 – Rank Effect with Over Different Holding Periods Individual Investor 1st Decile*Best

0.264 (4.28)

1st Decile*Worst

2nd Decile*Best

0.216 (3.92)

3rd Decile*Best 4th Decile*Best 5th Decile*Best 6th Decile*Best 7th Decile*Best 8th Decile*Best 9th Decile*Best 10th Decile*Best Stock x Date x HP Observations R2

Mutual Fund 0.054 (4.52)

1st Decile*Worst

0.081 (7.26)

2nd Decile*Worst 0.104 (2.82)

2nd Decile*Best 0.082 (8.96)

2nd Decile*Worst

0.071 (7.11)

0.168 (3.44)

3rd Decile*Worst

0.073 (1.94)

3rd Decile*Best

0.076 (6.60)

3rd Decile*Worst

0.089 (6.84)

0.139 3.152 0.117 2.953 0.083 2.321 0.097 2.653 0.084 2.692 0.055 2.265 0.036 (2.29)

4th Decile*Worst

0.072 2.152 5th Decile*Worst 0.055 1.839 6th Decile*Worst 0.056 1.936 7th Decile*Worst 0.056 1.887 8th Decile*Worst 0.071 2.396 9th Decile*Worst 0.072 2.491 10th Decile*Worst 0.070 (2.31) X

4th Decile*Best

0.073 9.000 5th Decile*Best 0.068 8.745 6th Decile*Best 0.078 9.816 7th Decile*Best 0.074 9.192 8th Decile*Best 0.081 8.669 9th Decile*Best 0.075 6.871 10th Decile*Best 0.028 (1.86)

4th Decile*Worst

1,048,549 0.278

0.073 (1.80)

1st Decile*Best

0.103 9.311 5th Decile*Worst 0.108 9.072 6th Decile*Worst 0.112 11.227 7th Decile*Worst 0.110 9.770 8th Decile*Worst 0.121 9.579 9th Decile*Worst 0.093 5.803 10th Decile*Worst 0.080 (4.26) X

15,603,394 0.085

This table presents a linear regression of sell on a best and worst dummy interacted with decile of holding period. Stock x Date x HP indicates a fixed effect for each interaction of cusip, date and decile of holding period. The top number is the coefficient, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account for the investors. For the investor data only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. Stocks are not included on the day that position is opened. Data covers January 1991 to November 1996.

22

Table IA.15 – Rank Effect with Controls for Past Performance Clustering Standard Errors by Account, Date and Stock Individual Investor Mutual Fund [1] Best Worst

[2] 0.157 (19.00) 0.107 (19.54)

2nd Best 2nd Worst Return*Gain Return*Loss Gain Return*Gain *√Holding Days Return*Loss *√Holding Days Variance *Gain Variance *Loss √Holding Days Observations R2

0.045 (2.86) -0.155 (-6.77) 0.037 (7.61) -0.002 (-3.44) 0.004 (3.43) 5.914 (1.86) -3.644 (-2.48) -0.002 (-7.79) 1,048,549 0.010

-0.002 (-0.20) -0.036 (-1.69) 0.029 (6.98) -0.001 (-2.46) 0.002 (2.53) 4.475 (1.93) -3.306 (-2.56) -0.002 (-8.25) 1,048,549 0.032

[3] 0.205 (20.09) 0.147 (19.69) 0.125 (16.19) 0.085 (14.18) -0.019 (-1.90) 0.004 (0.18) 0.026 (6.84) -0.001 (-1.77) 0.002 (1.82) 3.790 (1.94) -3.047 (-2.63) -0.002 (-8.72) 1,048,549 0.047

[4]

[5] 0.109 (11.33) 0.163 (11.96)

0.034 (6.66) -0.272 (-11.92) -0.013 (-3.72) 0.000 (-4.48) 0.008 (10.41) 4.971 (1.79) -2.557 (-1.85) -0.002 (-5.07) 15,603,394

0.024 (4.43) -0.242 (-10.80) -0.014 (-4.03) 0.000 (-3.49) 0.007 (10.39) 5.088 (1.84) -2.329 (-1.83) -0.002 (-5.06) 15,603,394

[6] 0.119 (11.72) 0.169 (12.04) 0.105 (12.18) 0.122 (10.25) 0.017 (3.07) -0.222 (-9.91) -0.014 (-4.08) 0.000 (-2.81) 0.007 (10.34) 5.111 (1.86) -2.039 (-1.72) -0.002 (-5.07) 15,603,394

0.005

0.006

0.007

This table presents marginal effects from logit regressions. The dependent variable is a dummy variable equal to 1 if a stock is sold. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase. Investor data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. Mutual fund data are from report dates from January 1990 to June 2010. A fund must hold at least 20 CRSP merged securities to be included in the analysis. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date, cusip and account for the investors and date, cusip and WFICN for the mutual fund data.

23

Table IA.16 – The Rank Effect with Every Possible Trade Date Included NOTE: Coefficients in percent (multiplied by 100)

Best Worst 2nd Best 2nd Worst Return*Gain Return*Loss Gain Return*Gain *√Holding Days Return*Loss *√Holding Days Variance *Gain Variance *Loss √Holding Days Observations R2

All

1 to 20 Days Since Trade

21 to 250 Days Since Trade

250+ Days Since Trade

[1] 0.187 (24.61) 0.076 (10.24) 0.088 (17.56) 0.005 (1.01) 0.069 (5.54) -0.044 (-1.18) 0.049 (10.88) -0.003 (-5.27) -0.002 (-1.15) 3.452 (1.22) 0.033 (0.09) -0.017 (-25.80) 47,764,044

[2] 0.379 (20.97) 0.177 (9.56) 0.214 (16.75) 0.049 (3.78) 0.146 (5.00) -0.351 (-4.17) 0.112 (11.10) -0.008 (-5.13) 0.014 (2.83) 4.704 (0.94) -0.030 (-0.04) -0.024 (-16.32) 19,737,357

[3] 0.131 (25.47) 0.073 (14.28) 0.059 (17.08) 0.020 (6.09) 0.086 (8.18) -0.065 (-3.07) 0.025 (8.00) -0.004 (-8.27) 0.000 (0.47) 2.833 (1.51) -3.953 (-4.10) -0.005 (-20.28) 23,964,672

[4] 0.030 (5.82) 0.017 (3.18) 0.011 (3.11) 0.010 (2.60) -0.014 (-1.27) -0.064 (-2.05) 0.004 (1.13) 0.000 (0.51) 0.002 (1.72) 5.677 (4.48) 0.356 (0.40) -0.002 (-5.20) 4,062,015

0.032

0.025

0.013

0.010

This table presents marginal effects from logit regressions. The dependent variable is a dummy variable equal to 1 if a stock is sold. Both sell days and days where nothing is sold and the market is open are included. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase. Days since last trade is the sell date minus the date of the last trade (either a buy or sell). Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Data covers January 1991 to November 1996. An investor must hold at least 5 stocks to be included in the sample. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

24

Table IA.17 – The Rank Effect by Year Year 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

Individual Investor Worst Best

0.136 (8.99) 0.151 (10.41) 0.156 (13.59) 0.140 (11.45) 0.139 (16.22) 0.153 (16.70)

0.208 (12.86) 0.248 (10.30) 0.223 (9.55) 0.216 (12.80) 0.186 (19.69) 0.181 (21.48)

Mutual Fund Worst Best 0.167 0.105 (4.37) (9.65) 0.161 0.104 (5.12) (6.42) 0.102 0.085 (2.71) (3.34) 0.123 0.122 (3.95) (6.88) 0.095 0.100 (3.59) (4.52) 0.182 0.149 (4.71) (7.42) 0.072 0.099 (3.36) (8.11) 0.091 0.084 (3.84) (6.27) 0.133 0.107 (6.14) (9.13) 0.168 0.084 (6.02) (7.63) 0.178 0.071 (7.04) (6.17)

Year 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

Mutual Fund (Cont.) Worst Best 0.144 0.093 (5.17) (6.95) 0.160 0.098 (11.05) (6.38) 0.087 0.135 (5.39) (6.33) 0.116 0.105 (7.22) (8.65) 0.120 0.105 (4.93) (6.91) 0.114 0.095 (6.23) (5.42) 0.158 0.127 (7.03) (7.49) 0.198 0.136 (7.40) (8.80) 0.213 0.103 (10.87) (5.68) 0.234 0.091 (8.33) (7.95)

This table presents marginal effects from logit regressions where the analysis is conducted separately for each year of data. The dependent variable is a dummy variable equal to 1 if a stock is sold. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase. Additional controls are 2nd Best, 2nd Worst, Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Investor data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. Mutual fund data are from report dates from January 1990 to June 2010. A fund must hold at least 20 CRSP merged securities to be included in the analysis. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

25

Table IA.18 – The Rank Effect By Investor Sex and Age

Best Worst 2nd Best 2nd Worst Return*Gain Return*Loss Gain Additional Controls Observations R2

Male

Female

Age: 40 and Under

Age: 41 - 48

Age: 49 - 58

Age: 58+

[1] 0.198 (17.79) 0.143 (16.30) 0.114 (15.62) 0.085 (13.81) -0.009 (-0.69) 0.035 (1.46) 0.032 (6.67) X 328,676

[2] 0.216 (9.56) 0.093 (4.01) 0.127 (8.53) 0.056 (4.61) -0.029 (-0.67) -0.005 (-0.05) 0.031 (2.83) X 36,841

[3] 0.188 (14.04) 0.119 (6.65) 0.114 (10.97) 0.085 (6.39) 0.018 (0.99) 0.064 (1.48) 0.048 (8.01) X 73,165

[4] 0.218 (7.94) 0.156 (9.31) 0.133 (6.40) 0.085 (6.26) -0.042 (-1.71) 0.109 (2.27) 0.036 (3.49) X 80,143

[5] 0.214 (14.23) 0.147 (7.55) 0.125 (11.72) 0.089 (7.31) -0.002 (-0.08) 0.026 (0.42) 0.026 (3.18) X 82,499

[6] 0.180 (11.89) 0.128 (9.82) 0.104 (14.46) 0.074 (9.39) 0.012 (0.71) -0.046 (-1.13) 0.024 (2.56) X 116,615

0.048

0.038

0.051

0.058

0.049

0.041

This table presents marginal effects from logit regressions. The dependent variable is a dummy variable equal to 1 if a stock is sold. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase. Days since last trade is the sell date minus the date of the last trade (either a buy or sell). Sex and Age are of the household head. Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

26

Table IA.19 – The Rank Effect by Income and Net Worth

Best Worst 2nd Best 2nd Worst Return*Gain Return*Loss Gain Additional Controls Observations R2

Less Than $50,000

Income $50,000 to $87,000

Less Than $75,000

Net Worth $75,000 to $175,000

$87,000+

$175,000+

[1] 0.187 (11.74) 0.105 (8.02) 0.110 (11.07) 0.059 (7.78) -0.012 (-0.41) -0.024 (-0.40) 0.026 (3.50) X 77,039

[2] 0.200 (9.29) 0.140 (9.57) 0.118 (6.88) 0.076 (6.53) -0.024 (-1.25) -0.071 (-1.70) 0.017 (2.14) X 89,133

[3] 0.197 (7.56) 0.158 (12.06) 0.132 (6.29) 0.097 (7.82) -0.013 (-1.06) 0.067 (2.37) 0.021 (2.78) X 211,974

[4] 0.181 (12.75) 0.136 (5.50) 0.107 (10.02) 0.067 (4.12) 0.041 (1.96) 0.022 (0.31) 0.029 (2.53) X 45,899

[5] 0.236 (6.74) 0.168 (7.37) 0.146 (4.84) 0.097 (4.45) -0.003 (-0.21) 0.041 (1.08) 0.015 (1.64) X 137,840

[6] 0.174 (12.62) 0.137 (12.44) 0.117 (11.64) 0.086 (11.59) -0.036 (-2.41) 0.011 (0.33) 0.025 (4.73) X 195,220

0.038

0.045

0.060

0.043

0.070

0.045

This table presents marginal effects from logit regressions. The dependent variable is a dummy variable equal to 1 if a stock is sold. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase. Days since last trade is the sell date minus the date of the last trade (either a buy or sell). Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

27

Table IA.20 – The Rank Effect Based on Returns in Excess of the Market

Best Worst 2nd Best 2nd Worst Market

All Gain (Market)

All Loss (Market)

[3]

[4]

[5]

0.214 (19.74) 0.140 (20.91) 0.122 (15.01) 0.089 (16.09)

0.117 (5.25) 0.030 (1.63) 0.088 (4.67) 0.040 (2.02)

0.101 (8.94) 0.053 (4.40) 0.042 (4.16) 0.032 (3.18)

-0.020 (-2.58) 0.048 (3.29) 0.007 (2.05) X 1,047,057

0.021 (0.42)

X 1,047,057

-0.010 (-1.23) 0.080 (5.53) 0.008 (2.44) X 1,047,057

X 6,369

X 21,852

0.047

0.045

0.046

0.010

0.014

[1] 0.202 (19.92) 0.145 (20.52) 0.123 (15.91) 0.084 (14.56) -0.016 (-1.00)

[2] 0.206 (19.06) 0.138 (21.08) 0.127 (15.54) 0.078 (14.42)

Best (Market) Worst (Market) 2nd Best (Market) 2nd Worst (Market) Return*Gain Return*Loss Gain

-0.019 (-2.73) 0.004 (0.24) 0.027 (7.97)

Return (Market)*Gain (Market) Return (Market)*Loss (Market) Gain (Market) Additional Controls Observations R2

0.134 (3.44)

This table presents marginal effects from logit regressions. The dependent variable is a dummy variable equal to 1 if a stock is sold. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase unless the variable has (Market) next to it indicating the variable is defined based on the return in excess of the market return since purchase. Market return is the CRSP value weighted index. Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. The All Gain (Loss) column limits the analysis to portfolios where all returns are above (below) the market return. Data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

28

Table IA.21 – Rank Effect for Individual Investors with Fixed Effects

Best Worst 2nd Best 2nd Worst Return*Gain Return*Loss Gain Additional Controls Account FE Stock FE Account x Date Stock x Date Observations R2

[1] 0.126 (31.55) 0.062 (16.84) 0.066 (26.41) 0.023 (10.44) 0.007 (1.09) -0.045 (-2.13) 0.028 (8.28) X X

[2] 0.174 (43.13) 0.108 (32.01) 0.106 (33.02) 0.062 (24.73) -0.031 (-4.43) 0.041 (2.10) 0.030 (9.36) X X

[3] 0.127 (32.73) 0.063 (17.81) 0.066 (26.71) 0.023 (10.77) 0.005 (0.67) -0.037 (-1.84) 0.033 (9.75) X X X

[4] 0.119 (27.51) 0.060 (14.44) 0.058 (21.63) 0.019 (7.43) 0.013 (1.77) -0.032 (-1.40) 0.031 (7.99) X

[5] 0.139 (23.06) 0.102 (17.49) 0.091 (18.94) 0.057 (12.44) -0.050 (-4.56) 0.180 (5.25) 0.037 (8.45) X

[6] 0.079 (10.72) 0.051 (6.80) 0.044 (7.80) 0.014 (2.57) -0.017 (-1.27) 0.089 (1.93) 0.045 (7.61) X

X 1,047,057

1,047,057

1,047,057

1,047,057

X 1,047,057

X X 1,047,057

0.067

0.053

0.082

0.119

0.678

0.767

This table presents results from linear probability models. The dependent variable is a dummy variable equal to 1 if a stock is sold. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Account FE indicates a fixed effect for each account, stock FE indicates a fixed effect for each CUSIP, Account x Date FE is a fixed effect for the interaction of account and date, Stock x Date FE indicates a fixed effect for the interaction of CUSIP and date. Return is the return since purchase. Additional controls are 2nd Best, 2nd Worst, Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

29

Table IA.22 – Rank Effect for Mutual Funds with Fixed Effects

Best Worst 2nd Best 2nd Worst Return*Gain Return*Loss Gain Additional Controls Account FE Stock FE Account x Date Stock x Date Observations R2

[1] 0.034 (7.88) 0.087 (15.76) 0.034 (10.70) 0.054 (12.78) 0.033 (8.80) -0.276 (-16.75) -0.013 (-5.31) X X

[2] 0.099 (12.94) 0.146 (14.44) 0.086 (13.74) 0.102 (12.04) 0.008 (1.82) -0.241 (-11.55) -0.018 (-6.30) X X

[3] 0.034 (8.28) 0.087 (16.16) 0.032 (10.88) 0.053 (13.07) 0.027 (7.80) -0.264 (-16.72) -0.014 (-6.25) X X X

[4] 0.041 (10.46) 0.110 (21.45) 0.039 (14.26) 0.073 (18.86) 0.034 (10.34) -0.226 (-12.72) -0.007 (-3.78) X

[5] 0.094 (11.69) 0.123 (12.25) 0.080 (12.51) 0.088 (10.78) 0.005 (1.05) -0.160 (-5.11) -0.016 (-6.15) X

[6] 0.037 (10.97) 0.077 (17.22) 0.034 (13.38) 0.051 (15.20) 0.029 (9.51) -0.143 (-11.56) -0.009 (-6.29) X

X 15,603,394

15,603,394

15,603,394

15,603,394

X 15,603,394

X X 15,603,394

0.135

0.026

0.141

0.326

0.108

0.389

This table presents results from linear probability models. The dependent variable is a dummy variable equal to 1 if a stock is sold. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Account FE indicates a fixed effect for each WFICN, stock FE indicates a fixed effect for each CUSIP, Account x Date FE is a fixed effect for the interaction of WFICN and date, Stock x Date FE indicates a fixed effect for the interaction of CUSIP and date. Return is the return since purchase. Additional controls are 2nd Best, 2nd Worst, Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Data are from report dates from January 1990 to June 2010. A fund must hold at least 20 CRSP merged securities to be included in the analysis. The top number is the marginal effect, and the lower number in parenthesis is the tstatistic. Standard errors are clustered by date and WFICN.

30

Table IA.23 – The Rank Effect Controlling for Tax-based Selling Based on Capital Gain or Loss Realized in a Tax Year 4th All Year Quarter Capital Gain*Best Capital Loss*Best Capital Gain*Worst Capital Loss*Worst

Additional Controls Observations R2

[1] 0.263 (23.97) 0.030 (3.83) 0.062 (9.49) 0.412 (37.28)

[2] 0.243 (22.39) 0.044 (4.60) 0.104 (12.57) 0.412 (28.95)

X 1,048,549

X 278,644

0.063

0.060

All Year

4th Quarter

[3] Capital Gain>$3,000*Best 0.268 (21.63) Capital Gain≤$3,000*Best 0.273 (23.92) Capital Loss<-$3,000*Best 0.055 (5.37) Capital Loss≥-$3,000*Best 0.016 (2.05) Capital Gain>$3,000*Worst 0.055 (7.96) Capital Gain≤$3,000*Worst 0.076 (10.07) Capital Loss<-$3,000*Best 0.419 (27.41) Capital Loss≥-$3,000*Best 0.416 (37.47) Additional Controls X Observations 1,048,549

[4] 0.242 (20.19) 0.264 (20.20) 0.055 (4.39) 0.036 (3.38) 0.089 (10.34) 0.142 (12.77) 0.395 (20.76) 0.431 (28.81) X 278,644

R2

0.063

0.060

This table presents the marginal effects from logit regressions of a dummy variable equal to 1 if a stock is sold on characteristics of the stock being held. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. Stocks are not included on the day that position is. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Capital Gain (Loss) is equal to the realized gain (loss) from purchase price of all securities in the portfolio from the beginning of the tax year up to that date of the tax year. The all year column includes all observations, while the 4th quarter column limits the analysis to sell days in the 4th quarter of the calendar year. Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Data covers January 1991 to November 1996. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

31

Table IA.24 – Alphabetical by Ticker Ordering Selling Last and Second First and Second to Last Ticker Ticker Only Only First Ticker

[1] 0.024 (3.28)

Last Ticker Stock by Day FE Observations

X 185,296

[2]

0.028 (3.54) X 185,088

All Tickers [3] 0.060 (10.55) 0.059 (10.59) X 1,016,954

Buying Last and Second First and Second to Last Ticker Ticker Only Only [4] 0.007 (2.11)

X 237,353

[5]

0.006 (1.47) X 237,120

All Tickers [6] 0.017 (6.40) 0.017 (6.27) X 1,396,848

This table presents regressions of a sell dummy (columns [1]-[3]) equal to 1 if a stock is sold or a buy dummy (columns [4]-[6]) on dummies based on ticker ordering and ticker by day fixed effects. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. Stocks are not included on the day that position is. First (Last) ticker is a dummy equal to one if the stock ticker is the first (last) ticker by alphabetical order in the portfolio. Data covers January 1991 to November 1996. The top number is the coefficient, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

32

Table IA.25 – Rank Effect for Buying Summary Statistics Best Worst 2nd Best 2nd Worst Return*Gain Return*Loss Gain Additional Controls Observations R2

[1] 0.006 (2.20) 0.038 (13.91) 0.009 (3.37) 0.032 (12.22)

Regression [2] 0.017 (6.24) 0.021 (9.80)

-0.015 (-5.61) -0.062 (-7.35) -0.010 (-8.87) X 1,440,981

[3] 0.022 (6.51) 0.030 (9.57) 0.017 (6.27) 0.022 (8.42) -0.017 (-6.46) -0.051 (-6.20) -0.009 (-7.98) X 1,440,981

0.041

0.046

Column [1] presents the difference in sale probability between the indicated rank stock and a stock not in the top or bottom two returns. Columns [2] and [3] present the marginal effects from logit regressions of a dummy variable equal to 1 if more of a stock that is already held is purchased. Only days where a stock is purchased and an investor holds at least 5 stocks are included in the sample. Stocks are not included on the day that position is opened. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio and 2nd Best (2nd Worst) is a dummy for the second highest (lowest) return. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Return is the return since purchase price, Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Data covers January 1991 to November 1996. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

33

Table IA.26 – Components of the Rank Effect using Various Measures of Extremeness [1] [2] Pos*(Return-Avg Return) -0.001 (-0.13) Neg*(Return-Avg Return) -0.001 (-0.18) -0.021 Pos*(Return-Median (-12.72) Return)/Std. Dev. 0.022 Neg*(Return-Median (8.93) Return)/Std. Dev. nd Best*(Best Return - 2 0.059 0.053 Best Return) (8.98) (9.70) Worst*(Worst Return -0.151 -0.169 nd (-14.55) (-16.16) 2 Worst Return) Best 0.188 0.237 (21.56) (21.74) Worst 0.118 0.139 (17.20) (17.58) Additional Controls X X Observations 1,048,549 1,048,549 R2

0.048

0.050

[3] [4] Best*(Return-Avg Return) -0.057 (-6.25) Worst*(Return-Avg Return) -0.031 (-4.61) -0.035 Best*(Return-Median (-13.68) Return)/Std. Dev. 0.017 Worst*(Return-Median (6.88) Return)/Std. Dev. nd Best*(Best Return - 2 0.099 0.071 Best Return) (11.07) (12.68) Worst*(Worst Return - 2nd -0.128 -0.176 Worst Return) (-11.93) (-15.43) Best 0.212 0.306 (24.82) (32.82) Worst 0.105 0.147 (14.58) (17.69) X X 1,048,549 1,048,549 0.049

0.050

Best*(Return-Avg Return)/Std. Dev. Worst*(Return-Avg Return)/Std. Dev. Best*(ReturnMedian Return) Worst*(ReturnMedian Return) Best*(Best Return 2nd Best Return) Worst*(Worst Return - 2nd Worst Best Worst

[5] -0.054 (-13.61) 0.030 (8.53)

[6]

-0.038 (-4.98) -0.024 (-2.68) 0.075 0.089 (12.85) (10.27) -0.184 -0.133 (-15.94) (-11.79) 0.369 0.203 (37.65) (23.20) 0.180 0.111 (16.34) (15.09) X X 1,048,549 1,048,549 0.050

0.049

This table presents the marginal effects from logit regressions of a dummy variable equal to 1 if a stock is sold on characteristics of the stock being held. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample. Stocks are not included on the day that position is opened or on days when all stocks are liquidated. Return is the return since purchase price. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio. Average return, median return and Std. Dev. (standard deviation) is the given measure for an investor on a given day. Pos is a dummy variable equal to one if the number in parenthesis is greater than 0, and Neg is a dummy equal to one when the number is less than or equal to zero. Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Columns [4]-[8] also include controls for 2nd best and 2nd worst return. Investor data covers January 1991 to November 1996. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

34

Table IA.27 – Components of the Rank Effect for Buying using Various Measures of Extremeness Pos*(Return-Avg Return)/SD Portfolio Neg*(Return-Avg Return)/SD Portfolio

[1] 0.005 (3.74) -0.007 (-6.74)

Pos*(Return-Med Return)

[2]

[3]

-0.013 (-2.95) -0.007 (-0.84)

Neg*(Return-Med Return) Best*(Best Return - 2nd Best Return)

[6]

[7] -0.009 (-6.14) 0.007 (4.82)

-0.031 (-9.31) 0.010 (1.45)

Best

X 1,440,981

X 1,440,981

X 1,440,981

0.042 (11.49) 0.039 (7.02) X 1,440,981

0.037

0.036

0.038

0.048

Worst

R2

[5]

0.012 (4.97) -0.061 (-13.14)

Worst*(Worst Return - 2nd Worst Return)

Additional Controls Observations

[4] -0.009 (-6.55) 0.007 (4.44)

0.027 (7.81) 0.032 (7.81) X 1,440,981

0.008 (3.75) -0.025 (-6.76) 0.020 (6.74) 0.026 (9.31) X 1,440,981

0.047

0.047

[8]

0.007 (2.95) -0.029 (-8.14) 0.039 (10.97) 0.034 (6.95) X 1,440,981

-0.038 (-10.81) 0.014 (1.98) 0.017 (7.00) -0.031 (-6.44) 0.024 (8.12) 0.026 (8.39) X 1,440,981

0.048

0.048

This table presents the marginal effects from logit regressions of a dummy variable equal to 1 if a stock is bought on characteristics of the stock being held. Only days where a stock is purchased are included and an investor must hold at least 5 stocks to be included in the sample. Return is the return since purchase price. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio. Average return, median return and SD (standard deviation) is the given measure for a portfolio on a given day. Pos is a dummy variable equal to one if the number in parenthesis is greater than 0, and Neg is a dummy equal to one when the number is less than or equal to zero. Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Columns [4]-[8] also include controls for 2nd best and 2nd worst return. Investor data covers January 1991 to November 1996. The top number is the marginal effect, and the lower number in parenthesis is the t-statistic. Standard errors are clustered by date and account.

35

Table IA.28 - Mutual Fund Fraction Sold and Liquidation by Rank Fraction Sold Best Worst Return*Gain Return*Loss Gain Additional Controls Observations R2

All Holdings 0.078 (13.55) 0.175 (16.00) 0.028 (4.78) -0.264 (-11.26) -0.007 (-2.88) X 15,603,394 0.044

Sell Only 0.023 (3.76) 0.141 (12.33) 0.000 (-0.07) -0.299 (-12.34) -0.023 (-4.31) X 6,068,983 0.123

Liquidate All Holdings 0.038 (10.39) 0.127 (14.91) 0.003 (1.41) -0.155 (-11.85) -0.011 (-4.95) X 15,603,394

Sell Only 0.001 (0.23) 0.237 (16.30) -0.047 (-7.46) -0.300 (-9.59) -0.027 (-4.44) X 6,068,983

0.063

0.090

This table presents a linear regression in the Fraction Sold column where the dependent variable is the number of shares sold on a report date divided by the number of shares held the previous report date. The Liquidate column presents the marginal effects from logit regressions of a dummy variable equal to 1 if all shares of a stock are sold. The All Holdings column includes all holdings while the Sell Only column includes only accounts on days where some positions are sold. Additional controls are Gain, Return* Gain, Return* Loss, Return*√Holding Days*Gain, Return*√Holding Days*Loss, Variance *Gain, Variance *Loss, and √Holding Days. Mutual funds must hold at least 20 stocks to be included in the sample. Stocks are not included on the day their position is opened. Standard errors are clustered by fund (wficn) and date. Best (Worst) is a dummy variable equal to 1 if the stock has the highest (lowest) return in the portfolio. Gain (Loss) is a dummy variable indicating a positive (non-positive) return. Data covers January 1990 to June 2010.

36

Table IA.29 - Mutual Fund Alphabetical Ordering by Company Name Selling Last and Second First and Second to Last Name Name Only Only First Ticker

[1] 0.009 (3.58)

Last Ticker Stock by Day FE Observations

X 256,835

[2]

0.001 (0.24) X 256,846

All Names [3] 0.050 (6.70) 0.044 (5.61) X 15,339,812

This table presents regressions of a sell dummy (columns [1]-[3]) equal to 1 if a stock is sold or a buy dummy (columns [4]-[6]) on dummy variables based on the alphabetical ordering by company name ordering and stock by day fixed effects. A fund must hold at least 20 stocks to be included in the sample. Stocks are not included on the day that position is opened. First (Last) name is a dummy equal to one if the stock name is the first (last) name by alphabetical order in the portfolio. Standard errors are clustered by fund (wficn) and date. Data covers January 1990 to June 2010. The top number is the coefficient, and the lower number in parenthesis is the t-statistic.

37

Table IA.30 – Optimality of the Rank Effect Panel A: All Holdings Comparison 1 Month 0.693 0.000 0.063

Sold: Worst Sold: Middle Sold: Best

1 Year -2.303 -1.030 2.100

Panel B: Matched Pair Not Sold: Everything Else Sold: Worst Matched Pair: Not Sold - Sold Not Sold: Everything Else Sold: Best Matched Pair: Not Sold - Sold

0.151 0.710 -0.559 (-4.34) 0.083 0.072 0.011 (0.12)

-0.084 -2.279 2.195 (4.04) -0.134 2.097 -2.231 (-5.59)

This table presents average returns in excess of the market (multiplied by 100) following sell days. A stock is Best (Worst) if it has the highest (lowest) return in the portfolio. Return is the return since purchase. 1 month returns are the returns for 20 trading days after a sell date and 1 Year is 252 trading days. Panel A presents the average excess return subsequent to sale for the indicated rank. Panel B presents the average difference in returns between an extreme position that is sold and an equal weighted portfolio of the middle ranked positions that are not sold. Data covers January 1991 to November 1996. Only days where a stock is sold are included and an investor must hold at least 5 stocks to be included in the sample.

38