[ Music ] >> All right, so in this video, we're going to look at what are some of the defenses for the kind of lackluster performance, if you will, at [inaudible] pricing model and predicted returns, at least over the, you know, kind of the long history from the 1920s till, you know, kind of the most recent century here. When you looked at performance from 1931 to 1965, capital asset pricing model was performing extremely well, predicting returns in a way that, you know, the -- it should, okay? If you look at, you know, kind of the, you know, more long time series, beta seems to only weakly predict stock returns and less so, you know, kind of since the mid1960s. So what's going on here? You know, advocates of the capital asset pricing model, I think their main defense would go something like this; the model maybe isn't perform as well -- performing as well as we would like because the market is mismeasured, okay? Theoretically when we talk about this, you know, kind of market, we are -- it should include all assets, not just the S&P 500, real estate should be in there, like a key asset for young folks who would be their human capital. We just simply proxy for the market portfolio. We aren't putting in all these, you know, kind of assets that really are in, you know, kind of a broad portfolio that people have. We just proxy with the, you know, stock market because we have, you know, good data on that. we have daily on the stock market but by using a proxy for the real market, this is inducing measurement error into the regression, and could buy us against, you know, kind of the CAPM performing as well as we would, you know, kind of expect it to. So mismeasurement of what the market portfolio really is by using this proxy of, you know, kind of the stock market, using this proxy as stock market maybe could explain somewhat why the capital asset price model doesn't perform as well as we would like it to. Another defense of the capital asset price model -- and just in a recently published paper in the Journal of Financial Economics back, I think, just in 2014, that the capital asset pricing model actually does perform pretty well on days where important macroeconomics, systematic news, is released, okay? So the authors of this are Savor and Wilson, 2014, Journal of Financial Economics piece. They basically, you know, do an analysis in spirit of Black, Jensen and Scholes and Fama and French that we've, you know, kind of talked about already in this module. They use daily data over the last 12 months to estimate betas, and then they form beta sort of portfolios over the next month, so they're kind of using a little different horizon for when they, you know, kind of form the betas, but basically the same idea. Use passed data to estimate betas for stock, rank stocks by their estimated beta, and then look into the -- look ahead and see do the stocks that have the higher beta typically have higher returns. But the rub that they're adding is they're looking, you know, does the CAPM predict well on days where there's a big news announcement about some macro, you know, kind of economic factor; that's what they're really bringing to the table, okay? What do we mean by macroeconomic news? Well, it could be news about inflation, employment, Federal Reserve, federal open market committee, interest rate decision, okay? So they're looking at days when those announcements occur. Now if you look at the trading days, these macro announcements are occurring about 13% of the time, roughly one, you know, kind of one out of eight, so you know, once every, you know, kind of two weeks or so. So that's 13% of the trading days, but when you look at the market risk premium, the performance of stocks, this -- these days when the macroeconomics news is announced, they account for about 60% of the equity risk premium during the year. So these are important days, 13% of the days but 60% basically of the stock market return occurring on these macro news announcements, okay? So let's look here -- and this is kind of a nice analysis to get things started here, where they look to show how important these announcement days are. Now remember, these announcement days here, showing, you know, kind of the dark bar here, only are 13% of the total trading days, but most of the term -- return of stock -- and here we have the excess return. This is in basis point, is occurring during days where you have this macroeconomic news. So macroeconomic news comes out, average excess return is a little over 10 basis points compared to about 1 basis point on non-announcement days. And Savor and Wilson compared this to, you know, kind of other turn of the month effects. So, you know, the first day of the month versus, you know, kind of not the first day and looking at returns during the month of January, the dark bar versus not the month of January, the light bar here. So a pretty big difference in the performance of stocks, rather there's a macroeconomic news coming out versus not, you know, kind of the difference here, much larger than, you know, if we're looking for example at returns in the month of January versus non-January, which is also a well-documented result. So let's look at some of the empirical results in their paper. Let's look at macro announcement days, which they call A days, A for announcement. What's the average excess return for 50 -- forget looking at 10, let's look at 50 beta sorted portfolios over the period 1964 to 2011, okay? So on our x axis, we have the capital asset pricing model beta, okay? And on our y axis, we have the average excess return in basis points here. And what do we see? An interesting distinction. The diamonds here, these are the returns of these 50 portfolios from the highest beta to the lowest beta here on announcement days. Here, you have the relation -- the highest beta to the lowest beta on non-announcement days. So on non-announcement days, the capital asset pricing model isn't, you know, kind of doing very well. There isn't really, you know, any relation if anything, is slightly negative, but you know, I won't make too much but that's basically no relationship between beta and the return. But when you look at the announcement days, when this macroeconomic news comes out, which should affect, you know, kind of stocks with systematic risk, you have a very clear relationship here. Firms that have higher beta, higher exposure to macro or systematic risks have bigger changes in their returns. Those with lower betas have lower returns on these announcement days, okay? And they're kind of, you know, very, very, very clear relationship. So when we look at the regression here of daily excess returns on, you know, kind of going back to just focusing on 10 beta sorted stocks over the 10 beta sorted portfolios over the period 1964 to 2011, let's recall the capital asset pricing model. So here, Savor and Wilson are basically doing an analysis pretty much, you know, similar to those done, you know, kind of more recently by Fama, Fama and French where we have 10 beta sorted portfolios, and we're just going to see if portfolio 10, which has the highest beta, has the -- has higher subsequent returns on average relative to portfolio 1, for example, which has the lowest betas. So remember the capital asset pricing model here, okay, predictions for the regression coefficients, the intercept term should be zero, okay? There's no intercept term predicted. There's no constant as part of this equation, so that's another way of saying when you do a regression, the prediction is the intercept term should be zero. The coefficient on beta, when we're regressing the return of the beta sorted portfolio on its average beta, the average beta, the stocks in the portfolio, the prediction is the coefficient on this regression should be just whatever the average excess return of the market was during that period, which in this case is 1964 to 2011, okay? And it's useful in terms of, you know, kind of daily average equity market premium, 10.5 basis points. So there's a clear prediction when we run this regression of the excess return of the beta sorted portfolio on its beta. This -- the coefficient on the beta should be 10.5 basis points if returns are being predicted by the capital asset pricing model in the way the capital asset pricing model says they should. Okay, let's get to the results. First, let's put up our regression equation for the beta sorted portfolios, we're regressing their excess return, okay, we're doing this, you know, on a daily rate, on the average beta of that portfolio. Prediction is higher beta, higher should be the return, okay? Intercept of this regression should be zero, there's no constant term here. Coefficient on beta should be this average excess market return during the period. They're using daily data so the average excess market return during this period was 10.5 basis points. What's the prediction -- or what are the results -- excuse me -- when we have -- just focus on announcement days so these are days where a macro announcement is occurring. There's, you know, federal -- U.S. federal government is announcing employment numbers, is announcing what inflation was, the Federal Reserve, federal open market committee is announcing a change in interest rates, so just focus in these announcement dates. Remember they're about 13% of all trading days. The intercept in this regression, 1.3 basis points, but not statistically significant different from zero, okay? What's the beta, what's the coefficient on beta? Usually we talk about [inaudible] when we say beta, we mean the coefficient, but in this case, what's the coefficient on beta? It's 9.2 basis point, 9.2, statistically significant result so it means a higher beta, higher returns earned by the portfolio, and the key is this 9.2 estimate is not statistically different from the capital asset pricing model prediction at 10.5. So in other words, on these macroeconomic news days, you can't reject the hypothesis that the capital asset pricing model predicts returns as it says it should, okay? So capital asset pricing model works, you know, basically very well in predicting which stocks go up the most on news when macroeconomic -- news concerning macroeconomic conditions. Now when we go to the non-announcement days, the N days, then the capital asset pricing model does not work well. And in fact, there's no relationship between the beta of the portfolio and the return of the portfolio on these non -- no announcement days, these N days. Coefficient here is, you know, statistically different from zero. So another way to think about this, if you're like trading stocks, if you think you have some insight about what's going to happen to, you know, kind of the macro economy, what the news is going to be, if you think the news is going to be very good about future economic development, you want to invest in stocks that have the highest beta because these regression results suggest the stocks with the highest beta are going to have the biggest, you know, kind of movement in returns when this macroeconomics news is released. So if you think good news is going to come out about the future economy, you want to be holding that day stocks that have the highest beta because they will have the biggest jump in price according to this regression analysis. So in conclusion, from A Tale of Two Days, days when macroeconomic news is released versus days where it's not, beta is strongly related to average returns as a CAPM would predict on the days when macroeconomic news is released. Kind of makes sense, right? Beta is a measure of the systematic risk or the macroeconomic risk of the stock. So on the days when this macro news comes, that's when the CAPM performs pretty well. But it is a puzzle. Why no relation between the stock beta and returns on non-announcement days, okay? So the authors of this study, A Tale of Two Days, speculate maybe the announcement days provide a clearer signal of, you know, systematic risks and expected future stock market returns to investors, therefore, you know, by getting this clear signal, they kind of trade according. And then that's, you know, kind reflected in the prices. So bottom line, there is kind of a tale of two days here, when macroeconomic news is being released, capital asset pricing model's pretty good at predicting returns on days outside of that, doesn't seem to work very well at all.
