Trading Strategies and Financial Models c 2012 Alexander Barinov 1 Multiple Choice Questions (10 questions, 3 points each) 1. Assuming that the momentum eect is mispricing, you would expect that a Recent losers will have positive returns around earnings announcements, and even more so if they were highly volatile before they started losing value b Recent losers will have negative returns around earnings announce- ments, and even more so if they were highly volatile before they started losing c Recent losers will have positive returns around earnings announcements, but less so if they were highly volatile before they started losing value d Recent losers will have negative returns around earnings announcements, but less so if they were highly volatile before they started losing value Losers continue to lose because they are still overpriced even given the initial loss. The overpricing is stronger for the stocks with high limits to arbitrage and will be corrected at earnings announcements. 1 2. If the value eect is explained by risk, then the value minus growth portfolio should a Have positive market beta b React positively to a sudden increase in dividend yield c Have high expected return in the periods of high GDP growth d Have negative CAPM alphas in the periods of increasing market volatility If value eect is an anomaly, it means that the market beta cannot explain why the returns to value rms and growth rms are dierent. Hence, the market beta of the value minus growth portfolio can be anything, it does not matter. If the value minus growth portfolio gains in response to a sudden increase in dividend yield (usually a signal of bad times), it is a hedge against recessions, and it is even more puzzling why it has a positive CAPM alpha. If the expected return (and, hence, the risk) of the value minus growth portfolio is high when GDP growth is high (good times) and low otherwise, it is a hedge again. But negative CAPM alphas when market volatility increases (usually during market downturns) clearly point towards the value minus growth port- folio being risky, which may explain its positive average CAPM alpha. 3. MOM is the portfolio that buys recent winners and shorts recent losers. If the regression of MOM on the excess return to the market yields MOM = 0.77 0.12 (MKT RF), what do you conclude? a One can earn 77 bp per month if one buys the winners and shorts the losers b One can earn 77 bp per month if one buys the winners, shorts the losers, and buys the market for $12 for each $100 one turns around in MOM c One can earn 77 bp per month if one buys the winners, shorts the losers, and shorts the market for $12 for each $100 one turns around in MOM d One can earn 77 bp per month if one shorts the winners, buys the losers, and buys the market for $12 for each $100 one turns around in MOM 2 You need to buy to cancel out the negative beta. You will not get 77 bp if you do not, you will get 77 bp less 0.12 times the market risk premium. (d) has the momentum strategy backwards. 4. The 3-0-3 momentum strategy suggests buying winners and shorting losers based on their return in the previous 3 months and holding to the positions for 3 months after opening. Assuming that all positions are closed at the end of the 3 months holding period, what is the highest level of one-way trading cost at which the strategy will still be protable, if its average abnormal annual return is just a bit over 15%? a 0.9375% = 15% 4 4 b 1.875% c 3.75% d 7.5% When you open and close the positions in the momentum strategy, you have to pay the one-way trading costs 4 times: when you buy winners, when you sell them back, when you short losers, when you buy them back. Also, if the holding period is 3 months, you have to do it 4 times a year. If the one- way trading costs are above 0.9375%, trading costs per year, > 0.9375% 4 4, beat the return to the strategy, 15%, and the strategy is not protable on the after-cost basis. 5. Mispricing or apparent mispricing should be stronger for a Volatile rms, large rms, rms with low Roll measure b Firms with high Amihud measure, rms with high bid-ask spread, rms with high short interest c Low volatility rms, rms with low Amihud measure, rms with high prob- ability to be on special d Small rms, rms with high residual institutional ownership, illiquid rms 3 We are looking for the rms with high limits to arbitrage (mispricing) or high trading costs (apparent mispricing). In (a), both large rms and rms with low Roll measure (high bid-ask spread) do not t this description. In (c), low volatility rms have low limits to arbitrage and rms with low Amihud measure (low price impact) have low trading costs. In (d), rms with high IO have low limits to arbitrage. 6. If the new issues puzzle is risk, then, controlling for the market return, the correlation between the realized returns to new issues and the returns to the futures on the Treasuries should be a Negative b Positive c Positive in expansions and negative in recessions d Impossible to tell New issues have negative CAPM alphas, thus, a risk-based explanation should show that they have low risk, e.g., that they lose less value in reces- sions than what the CAPM predicts. Treasuries usually gain during recessions (Treasury bill rate goes down in recessions, and Treasury prices go up), and so do futures on them. Hence, controlling for the market return, the correlation between the realized returns to new issues and the returns to the futures on Treasuries should be positive. 4 7. Assuming that the value eect is caused by risk, you would expect the realized returns to the HML factor to be a Positively correlated with future labor income and positively cor- related with future Treasuries yield b Negatively correlated with future labor income and positively correlated with future Treasuries yield c Positively correlated with future labor income and negatively correlated with future Treasuries yield d Negatively correlated with future labor income and negatively correlated with future Treasuries yield Risk is losing money just before things turn south - i.e., labor income drops and Treasury yields (and expected ination) decrease. 8. You are creating a multifactor model with the momentum factor. The momentum factor is the portfolio long in winner rms and short in loser rms. You can interpret this model as the ICAPM a Only if momentum is mispricing b Only if momentum is risk c The ICAPM does not allow the use of the momentum factor d You can use the momentum factor in the ICAPM irrespective of whether you think momentum is risk or mispricing ICAPM is a risk-based model, in contrast to APT, which just assumes that the factors are there, and other multifactor models, which may ask the question if the LHS variable is simply related to existing anomalies. All ICAPM factors have to be risk. 5 9. If the new issues puzzle is mispricing, then it should be stronger for the rms with a Low institutional ownership and high short ratio b Low institutional ownership and low short ratio c High institutional ownership and high short ratio d High institutional ownership and low short ratio New issues have negative alphas, and you have to short them to make prot and to arbitrage away the anomaly. In the subsample of rms that are costly to short no one is able to short new issues (or no one nds it protable) and the new issues puzzle persists. In the subsample of rms that are easy to short, everyone shorts new issues, they stop being overpriced, and the new issues puzzle disappears. Short sale costs are high if either the demand to short is high or the supply of shares to short is low. Demand is proxied for by short ratio, and supply is proxied for by institutional ownership. Shares that are hard to short normally have high short ratio and/or low institutional ownership, and the new issues puzzle should be stronger for these rms. 10. If the idiosyncratic volatility discount is mispricing, then a Firms with low idiosyncratic volatility should be more mispriced if they also have high probability to be on special b Firms with low idiosyncratic volatility should be more mispriced if they also have low probability to be on special c Firms with high idiosyncratic volatility should be more mispriced if they also have low probability to be on special d Firms with high idiosyncratic volatility should be more mispriced if they also have high probability to be on special 6 The idiosyncratic volatility discount refers to the positive alpha of the port- folio that buys low volatility rms and shorts high volatility rms. Firms with high volatility have negative alphas and need to be shorted. If investors can- not short them due to high shorting fees, their negative alphas will persist and they will remain overpriced. The high volatility rms that have low shorting fees will be actively shorted and will stop being overpriced. Probability to be on special is the probability of having the shorting fee so high that it exceeds the risk-free rate. Hence, high probability to be on special means high shorting fees, and the high volatility rms with high probability to be on special will be more overpriced. 2 Short Questions (5 questions, 10 points each) 1. High turnover of the stock (trading volume divided by market capitalization) is usually interpreted as either high uncertainty about the stock value or high liquidity of the stock. Explain how, under the hypothesis that the post earnings announcement drift is mispricing, each interpretation of turnover would impact your prediction about the cross-sectional relation between turnover and the strength of the post earnings announcement drift. If PEAD is mispricing, then it should be stronger if limits to arbitrage are higher. High uncertainty is a limit to arbitrage - it can be dangerous to hold the stocks you do not understand, even if you are after a regularity in returns that requires no analysis of separate companies. If turnover is uncertainty, then PEAD should be stronger for high turnover stocks. High liquidity means low limits to arbitrage (the stock is easy to buy and sell). If turnover is liquidity, then PEAD should be weaker for high turnover rms. Overall, the prediction on the relation between the PEAD strength and turnover is ambiguous. Empirically, it is positive, so turnover as uncertainty wins. For your use, you can separate turnover into average and unexpected - average is liquidity, unexpected is uncertainty. 7 2. Investment-to-capital ratio is known to be higher during expansions (rms invest a lot if demand is high). What is the expected sign of the correlation between expected market return and the lagged value of the aggregate investment-to-capital ratio to be? Explain. High investment-to-capital ratio means expansion, when risks are low and expected returns are also low. Therefore, high investment-to-capital ra- tion should predict low expected market returns. That is, the correlation has to be negative. 3. You are trying to explain the new issues puzzle using conditional CAPM. What will be your hypothesis about the relation between the CAPM beta of IPOs and lagged values of GDP growth and unemployment rate? Explain. IPOs have negative CAPM alphas. Therefore, they have to be less risky than what the CAPM thinks. In the conditional CAPM, the only form of low risk is the decrease in the CAPM beta during recessions. GDP growth is low during recessions, hence it has to be positively correlated with the CAPM beta of IPOs to explain the new issues puzzle. The unemployment rate is high in recessions, hence it has to be negatively correlated with the CAPM beta of IPOs. 4. You have noticed that the zero-investment portfolio buying rms that do not invest a lot and shorting rms that do invest a lot has positive and signicant CAPM alpha. The next regression you ran was the two-factor model with the market factor and the HML factor, and in this model the investment-based arbitrage portfolio turned out to have an insignicantly negative alpha. What do you conclude if you believe that the value eect is risk? How does the conclusion change if you believe that the value eect is mispricing? If the value eect is risk, than HML (which is essentially the value minus growth portfolio) is a risk factor. The fact that adding it to the market model eliminates the alpha of the investment strategy means in this case that the CAPM alpha of the investment strategy is not abnormal return, but rather a fair compensation for the value risk. 8 If you believe that the value eect is mispricing, you would conclude from the insignicant alpha of the two-factor model that the investment anomaly is just another reincarnation of the value eect. 5. The Roll measure of Delta Airlines (DAL) is 1.2%, the Amihud measure of Delta Airlines is 0.009%. What is the dollar cost of buying $50M worth of Delta shares in one day? Assume that the true price is at the midpoint and the price impact starts at the ask. You lose half of the bid-ask spread when you buy every share of DAL. The Roll measure measures the bid-ask spread. Hence, you lose 0.6% $50M = $300K on the bid-ask spread. Then the price impact happens. On the rst lot, you lose nothing. By the time you trade the last lot, the price has gone up by 0.009% 50 = 0.45% (0.009% is the Amihud measure). On average, you lose half of this number, or, in dollars, 0.225% $50M = $112.5K. The total loss of the one-way trade is $300K+$112.5K=$412.5K. 3 Long Question (4 parts, 5 points each) This question refers to the ICAPM, estimated for Vanguard International Value (VTRIX) fund using the data from January 1999 to December 2009. The numbers in brackets are standard errors. Ret t RF t = 0.35 (0.24) + 0.88 (0.05) (MKT t RF t ) + 1.21 (6.55) TB t 5.86 (2.88) DIV t Looking at the coecient on the change in the dividend yield, does the fund look riskier than what its market beta implies? Explain. Dividend yield is high in recessions, so an increase in dividend yield is bad news. The negative slope on DIV suggests that, controlling for the market risk, the fund responds to the bad news by posting negative return, which means that the fund is riskier than what its market beta implies. (Note also that the coecient is statistically signicant). 9 If dividend yield increases by 1% and the expected ination decreases by 3%, by how much will the fund beat/trail the CAPM? If dividend yield increases by 1%, the fund will post -5.86% loss on top of the loss coming from the market decline that normally accompanies increases in dividend yield. If expected ination decreases by 3%, the Treasury bill rate also decreases by 3%. The fund will respond by posting 1.21 3% = 3.63% loss on top of the loss coming from the market decline that normally accompanies decreases in the Treasury bill rate. Hence, if dividend yield increases by 1% and the expected ination decreases by 3%, the fund will underperform the CAPM by 5.86%+3.63%=9.49%. (Note: 5.86 and 1.21 in the solution are slopes from the regression above). Your colleague suggests that the Vanguard fund is a good investment, because the ICAPM above estimates that its alpha is positive. Give two objections. One obvious objection is that the alpha is not signicant (0.35/0.24=1.462). The second more subtle objection is that the dividend yield on the market portfolio is not a return to a tradable asset, hence, the alpha does not have to equal zero and can be anything even if the fund yields no abnormal return. Another colleague suggests that you throw in the change in GDP as the fourth factor. What sign of the slope on GDP growth would suggest that the fund is even riskier than what you would think looking at the regression above? Low values of GDP growth mean recessions. Risk means losing money in recessions, i.e., having low return when GDP growth is low. Hence, if the fund is even riskier than what you would conclude from the regression above, the slope on GDP growth should be positive. 10