FIN 620 - Summaries of all 25 required readings in FSS 2021 PDF

Preview

Summary

Da et al. (2011) - In Search of AttentionMain contributions of the paper: Da et al. (2011) propose a new and direct measure of retail investor attention using search frequency in Google (Search Volume Index) Their measure is correlated with but different from existing attention proxies and captures ...

Description

Da et al. (2011) - In Search of Attention Main contributions of the paper: 1. Da et al. (2011) propose a new and direct measure of retail investor attention using search frequency in Google (Search Volume Index) 2. Their measure is correlated with but different from existing attention proxies and captures investors’ attention in a more-timely fashion 3. They test and confirm the attention theory of Barber and Odean (2008) 4. Their proxy also contributes to explaining short-term and long-term patterns in IPO returns Existing Attention Measures as Starting Point Attention proxies: ▪ Extreme one-day returns ▪ Abnormal trading volume ▪ News coverage ▪ Advertising expense ▪ Upper price limits

Barber and Odean (2008) X X X

They have weaknesses their implicit assumptions: ▪ Investors pay attention to extreme returns/turnover and to coverage in the news. However:  Return and trading volume can be driven by factors unrelated to attention  News result in attention only if they are read The main attention measure: Abnormal Google Search Volume Index (ASVI) for ticker symbols ▪ SVI is a direct measure of retail investor attention ▪ In contrast to existing proxies it measures revealed attention and is weakly correlated with other measures ▪ ASVI: ASVIt=log(SVIt) – log[Med(SVIt-1, … ,SVIt-8)] ▪ Weekly search volume index for individual stocks from Google Trends ▪ Identification of a stock via ticker: − avoids problem of multiple reference names (less ambiguous) − captures interest in financial information ▪ Two problems with names: − Investors may search name for reasons unrelated to investing (e.g., Apple) − Investors may search the same firm using variations of its name ▪ Noisy tickers (e.g., GPS = Gap, Inc.) fagged ▪ SVI based on the ticker ▪ SVI based on company name: Name_SVI ▪ SVI based on main product of company: PSVI (based on advertisements) ▪ Data downloaded using a web crawling program ▪ SVI measures investors’ attention continuously over the year while news coverage of a typical firm is sporadic

▪ If search volume is triggered by news event, SVI carries additional information about the amount of attention generated SVI and Other Attention Measures Empirical approach:  Vector Autoregression (VAR) is used to measure the relationship between SVI and other attention measures Results:

 SVI leads other three attention proxies: positive coefficients suggest that SVI captures investor attention in more timely fashion than turnover, extreme returns or news.  Investors continue to pay attention to stocks after extreme returns

Whose attention does SVI capture?  SVI Captures Attention of Retail Investors Empirical approach: ▪ Data of retail orders and trades per market center  SEC Rule 11Ac1-5 (Dash-5) and Rule 605 requires every market center to report on a monthly basis  Market centers attract different retail investors given their information level ▪ Dash-5 includes:  # shares traded, # orders received, and various dimensions of execution quality  Disaggregated trading statistics: (100-499); (500-1,999); (2,000-4,999); (5,000-9,999) shares ▪ Compute monthly changes in orders and turnover and relate them to monthly changes in SVI ▪ Control for alternative attention proxies and stock characteristics Results: Interpret panel A: A 1% increase in SVI leads to a 0.0925% increase in individual orders and a 0.0919% increase in turnover Results in regression (3) and (4) are similar to (1) and (2) → Pool data (Panel B)

Interpret panel B: Effect stronger at market system that most likely attracts order fow from less informed individual investors (Madoff) SVI and Price Pressure ASVI is a direct measure of retail attention, we can directly test the price pressure hypothesis of Barber and Odean (2008): Large ASVI results in increased buying pressure that pushes stock prices up temporarily They test it in the context of Russell 3000 stocks and then in the context of IPOs: 1. Finding on Russell 3000 stocks -

-

Strong evidence of positive price pressure following an increase in individual attention as measured by ASVI a significant and negative coefficient on the interaction term between Log Market Cap and ASVI  Effect of ASVI is stronger among small stocks because they are more infuenced by individual investors Positive coefficient of the interaction between ASVI and Percent Dash-5 Volume (which measures retail investors’ trading)  a stronger price increase among stocks traded mainly by retail investors  further confirm that the channel in which ASVI affects stock price is mainly through retail investors

1.1. A rational explanation for the effect of ASVI on stock price: Positive effect of ASVI on price could also simply refect positive fundamental information about the firm. For example, suppose a company announces an innovation in its product to which consumers react positively  ASVI increases This alternative explanation implies that there could be another factor driving both ASVI and stock price  potentially reject the effect of ASVI on price

If this information story is true, we would expect an even larger positive coefficient on APSVI, which subsumes the predictive power of ASVI when we include APSVI in the regression.  not confirmed in the regression that puts ASVI and APSVI together! 1.2. Test the long-term reversal If ASVI drive prices through attention, this is irrational  we should expect a long-term reversal Test result: ASVI seems to be the only measure of attention that predicts both the initial price increase and subsequent long-run price reversal  confirm the prediction above! 2. Finding on IPO sample Two stylized facts about IPO returns: -

IPOs on average have large first-day returns IPOs exhibit long-run underperformance

Hypothesis: Since it is usually difficult to short-sell IPOs, buying pressure from retail investors can contribute to higher first-day returns Methodology: -

measure retail attention prior to the IPO using ASVI there is no ticker widely known to the public prior to the IPO  use company name provided by SDC

Results: (1) There are significant changes in SVI around the time of the IPO: -

significant upward trend in SVI starting 2 to 3 weeks prior to the IPO week, followed by a significant jump in SVI during the IPO week - SVI reverts to its pre-IPO level 2 to 3 weeks after the IPO  Consistent with IPO returns pattern! (2) Relation between increased attention prior to IPO and first-day IPO return: -

IPOs with low ASVI during the week prior to the IPO: return = 10.90% IPOs with high ASVI during the week prior to the IPO: return = 16.98% The difference is statistically significant at the 1% level Both formal tests (ASVI standalone and with controls) show significant positive effect of ASVI on IPO first-day returns  Consistent with the attention-induced price pressure hypothesis Explanation: Since it is usually difficult to short-sell IPOs, buying pressure from retail investors can contribute to higher first-day returns.  Implication from this explanation: when the price pressure due to excess retail demand disappear, stock prices eventually reverse  long-run underperformance 3. Alternative Interpretation for ASVI and IPO returns:

Market participants have an expectation of IPO first-day returns and that they search a lot (a little) prior to the IPO when they expect first-day return to be high (low)  reverse causality issue: higher expected first-day returns cause higher ASVI  Test this “anticipation hypothesis”: Directly measure market expectations of first-day returns using IPO SCOOP: -

Include this measure to the regression of first-day returns on ASVI  estimate of ASVI changes little and remains significant!

-

Regress post-IPO returns on IPO SCOOP ratings and its interaction with first-day returns  none of them is significant

 No evidence for anticipation hypothesis!

Odean (1998) – Are Investors Reluctant to Realize Their Losses? MAIN IDEA Odean tests the presence of disposition effect for individual investors, and then shows that it cannot be explained by rational reasons as the willingness of rebalancing portfolio or avoiding the trading costs of low-priced stocks. He rather finds an explanation in prospect theory and a mistaken belief (of investors) in mean reversion of the stocks. Disposition effect is problematic because it leads to a suboptimal behavior for taxable investments. For tax purpose, investors should keep their winners and sell their losing investments.

Disposition effect definition: Investors have the tendency to sell shares whose price has increased, while keeping assets which have dropped in value. Therefore, investors tend to sell the winners too early and hold the losers too long. According to Odean, it is possible to explain this effect with prospect theory. People behave as if maximizing an S-shaped value function.

Prospect theory as an explanation for Disposition effect: Relative to reference point (Odean uses the purchase price as proxy), an investor in the gain domain is risk averse and have more tendency to sell the stock; unless he updates his reference point and change to the loss domain. In the loss domain: an investor is more risk seeking and tends to hold the stock until it reaches gain domain; unless he updates the reference point and decide to realize the loss.  The aim of the study: to find whether investors sell the winners too soon and hold the losers too long. Furthermore, he investigates tax-motivated trading in December.

Data: Trading records from 1987 to 1993 for 10000 accounts (randomly selected) at a nationwide discount brokerage house.

In case of multiple buys or sells of the same stock in the same account and in the same day, they are aggregated. He obtains information on splits, dividends and price data from the 1993 Center for Research in Security Prices (CRSP).

Methodology 

It is not sufficient to look at how many stocks are sold for gains or losses



It is necessary to look at the frequency with which investors sell winners and losers relative to the possibility to sell each.

 Constructs for each date a portfolio of securities which purchase date and price are known. Determine if a stock is sold for a loss or a gain: 

Every day a sell take place in a two-or-more-stock portfolio  realized gains/losses



Compare the selling price to the stock’s purchase price for each stock.



Paper (unrealized) loss or gain: stocks that are in the portfolio at the beginning of the day and that have not been sold. To know if it is a paper gain or loss, compare its high and low price for that day to its average purchase price: 1. If both are below, it is considered paper loss 2. if both are above it is considered paper gain. 3. Otherwise, it’s not considered  neither paper gains nor losses

Measures the disposition effect: -

Proportion of Gains Realized (PGR) = �������� �����/(�������� ����� + ���������� ���

-

Proportion of Losses Realized (PLR) = �������� Losses/( �������� Losses + ���������� ������)

-

Disposition Effect = PGR – PLR

-

PGR and PLR are aggregated over all investors, i.e. Odean assumes paper sales/losses and realized gains/losses to be independent within and across investors

Define reference point: −

For testing the disposition effect, it’s necessary to consider a reference point, according to which it’s possible to specify gains and losses.

−

Possible choices for reference point: average purchase price, highest purchase price, first purchase price, most recent purchase price  used for robustness check  Findings: same results for each alternative

−

Odean reports results for average purchase price.

−

Dividends are not included (he finds that findings are not affected by the exclusion of commissions and dividends.

Main hypotheses: Hypothesis 1: PGR > PLR (for the entire year) Hypothesis 2: PLRDecember – PGR PLRDecember > PLRJanuaryNovember – PGRJanuaryNovember

Main Results: 

For both Hypothesis 1 and Hypothesis 2 the null hypotheses can be rejected PGR and PLR for the entire Data Set Entire Year

December

Jan.–Nov.

0.098 0.148 —0.050 —35

0.128 0.108 0.020 4.3

0.094 0.152 —0.058 —38

PLR PGR Difference in proportions t-statistic



These tests count each sale for a gain, sale for a loss, paper gain on the day of a sale, and paper loss on the day of a sale as separate independent observations.

 This independence assumption will not hold perfectly 

To gain some perspective into how critical the independence assumptions are to the primary finding  look at an alternative test

Alternative test: on account level Assumption: independence exists only at account level  proportions of gains and losses realized in each account are independent of those realized in other accounts. Methodology: 

PGR, PLR, and PGR – PLR are estimated for each account



Sale of a stock is only counted if no sale has been previously counted for that stock in any account within a week before or after the sale date  no two sales of the same stock within a week of each other are counted



Similarly, no two unrealized paper losses/gains of the same stock within a week of each other are counted



Difference to the main test: -

Main test: implicitly weights each account by the number of realized and paper gains and losses in that account

-

Alternative test: weights each account equally  ignore the fact that accounts with more transactions provide more accurate estimates

Result: Null hypothesis that the mean of PGR – PLR is less than or equal to zero is rejected with a tstatistic of 19. An alternative specification: Calculate PGR and PLR in terms of number of shares traded and potential number of shares traded. Methodology: PGR and PLR can be calculated first for each account and then the mean of PGR – PLR is calculated.  Accounts are weighted equally

Result: Null hypothesis that the mean of PGR – PLR is less than or equal to zero is rejected with a tstatistic of 18.  when test is done for PGR and PLR based on shares rather than trades, the results are virtually unchanged.

Rational explanations: 1. Transaction Costs: Investors may be reluctant to sell losers because they want to avoid the higher costs of low-priced stocks.  Even reporting PGR and PLR for different price ranges and return ranges, winners are always realized at a higher rate than losers. Also comparing winners and losers with the same magnitude and same trading costs, investors still prefer to sell winners and hold losers even when trading costs are the same.  Another way to test this explanation: It implies that losses are realized more slowly due to the higher transaction costs  look at the rates at which investors purchase additional shares of stocks they already own. Methodology: Calculate PGPA (proportion of gains purchased again) and PGPA (proportion of Losses purchased again):  

PGPA = (Gains purchased again)/(Gains purchased again + gains potentially purchased again) PLPA = (Losses purchased again)/(Losses purchased again + losses potentially purchased again)

Hypothesis: PLPA < PGPA if investors avoid the higher transaction costs of low priced stocks. Result: PLPA > PGPA  more consistent prospect theory (investors are more risk seeking in the loss domain) than with the Transaction cost hypothesis! 2. Portfolio rebalancing: Investors may sell winners and hold losers for rebalancing purpose.  Sales for rebalancing purpose would sell only a portion (but not all) of shares of winning stocks.  To eliminate trade that may be motivated by rebalancing, calculate PGR and PLR using only sales of entire position.  Result: Disposition effect still holds. PGR and PLR when the entire position in a stock is sold PLR PGR Difference in proportions t-statistic

Entire Year

December

0.155 0.233 —0.078 —32

0.197 0.162 0.035 4.6

 In case of sales for rebalancing purpose, a sale is usually followed by new purchases  only consider sales for which there are no purchases into a portfolio in the following three weeks.  Result: Disposition effect still holds. PGR and PLR when no new stock is purchased within three weeks of sale PLR

Entire Year 0.281

December 0.391

PGR Difference in proportions t-statistic

0.449 —0.168 —36

0.366 0.015 1.6

3. (Private) information about subsequent portfolio performance: Investors could expect the losers to outperform the winners.  Compute the excess returns for periods following the sale of winning stocks sold to excess returns of those with paper losses  Result: losers do not outperform winners in the future  this is a mistaken (so not rational) belief Ex post returns

Average excess return on winning stocks sold Average excess return on paper losses Difference in excess returns ~ p-values!

Performance over Next 84 Trading Days

Performance over Next 252 Trading Days

Performance over Next 504 Trading Days

0.0047

0.0235

0.0645

—0.0056

—0.0106

0.0287

0.0103 ~0.002!

0.0341 ~0.001!

0.0358 ~0.014!

Behavioral explanations 1. Mistaken belief in mean reversion: losers outperforming winners 2. Prospect theory 

Reference Point: Past prices



Investors are risk averse in the gain domain and risk seeking in the loss domain -

If a stock happens to be traded in the gain domain, investors tend to realize their gain and get the money to their account; because they are risk averse and don’t want to risk that they might end up in the loss domain

-

In the loss domain, investors are risk seeking. Hence, they don’t want to realize the losses while they still have the chance to end up in the gain domain  hold the stock

Frazzini (2006) - The Disposition Effect and Underreaction to News Post Earnings Announcement Drift (PEAD) Stocks of firms with unexpectedly good (bad) earnings tend to drift upwards (downwards) in the period (up to 3 months) after the earnings announcement. Disposition effect is the tendency of investors to ride losses and realize gains. Underreaction definition: New fundamental information slowly refected to prices, resulting in subsequent drift in the same direction. ABSTRACT: Main goal: Tests whether the “disposition effect induces “underreaction” to news and PEAD.

Methodology: use data on mutual fund holdings to construct a new measure of reference purchasing prices for individual stocks Main results:  Post-announcement price drift is most severe whenever capital gains and the news event have the same sign.  The magnitude of the drift depends on the capital gains (losses) experienced by the stock holders on the event date.  An event-driven strategy based on this effect yields monthly alphas of over 200 basis points. THEORY & HYPOTHESIS DEVELOPMENT 1. How Disposition Effect Leads to Underreaction and PEAD ■ Disposition effect may affect the supply of stocks: If stock prices rise above the reference point (gain overhang):    

Investors will be more willing to sell Supply of stocks increases Temporary downward pressure on prices Positive drift in stock prices (high future returns)

If stock prices fall below the reference point (loss overhang):  Investors...