Let's start our searching for Alpha here. Where are we going to look first? How about mutual funds and local investments? Our mutual fund investors more informed about the stocks of companies located in their backyard relative to those that are located farther away. Maybe for some types of firms, information investors nearby the firm may simply have an informational edge, just knowledge of the local economic trends, networking opportunity with local executives, maybe there is some geographic advantage in terms of information dissemination. Or maybe local investments don't outperform at all if they're simply motivated by a familiarity bias on the part of the mutual fund manager. It's certainly an empirical question to see how do local investments of a mutual fund manager perform relative to their more distant investments here. Coval and Moskowitz did a very influential study looking at the geography of mutual fund investments, and they examined quarterly mutual fund holdings. We talked just previously in the module about the disclosure requirements of mutual funds. They look at those quarter Lee mutual fund holdings of actively managed equity funds from 1975-1994. They define a holding is local to the mutual fund if the firm headquarters is located within 100 kilometers of the firm's headquarters. Let's think about this analysis here. Suppose some firm relevant information is first available locally, is there any useful information to be had from observing what stocks are not held by the local mutual funds? If one exists, are mutual funds more likely to have a local informational advantage on the stocks of small firms or large firms? These questions are set up to think how can we design our empirical tests to look is there really an informational advantage tied to geography. First question, is there any useful information by looking at what stocks are not held by local mutual funds? Then second, are mutual funds more likely to have an informational advantage on the stocks of small firms or large firms? Let's think about that and then I'll give you my take. What did you think? I think the natural hypothesis is if there's a local informational edge, then there's actually useful information, not only in what local companies are the mutual fund holdings. What Boston firms are being held by Boston mutual funds. But it's maybe also useful what Boston companies are the Boston mutual funds avoiding. That might be useful information as well. Just like we discussed with individual investors, when we're looking at mutual funds we may also think that the local informational advantage is going to be stronger for the stocks of smaller firms relative to large firms where there's just so much analysts and national media and international media attention. Coval and Moskowitz, they study the geography of mutual fund investments. They examine performance of a mutual funds local stock holdings relative to more distant stock holdings. They examine the performance of local stocks not held by a mutual fund as well, and also, they examine the performance of a local stock holding that is increased in size, so you're investing more in this stock versus those that have fallen in size as measured by the number of shares. They look at this performance, they're basically measuring the performance of the holding starting both from the quarter-end date, as well as starting three months after the quarter-end date. They'll look at the performance of the mutual funds, local investments at two points at time. They're going to consider how the performance of the stocks held by the local mutual fund varies by firm size, just as we suggested would be reasonable tasks to do in our pause, think, and answer a question, and they're going to report underneath the coefficients. They're going to report t-statistics and parentheses. They're going to report the coefficients are what are the various annual returns below that are the t-statistics and parentheses. Remember from our statistics primer, let's look for t-statistics with a magnitude of at least two to indicate well, this seems to be statistically different than zero conventional levels. Let's look here at their first analysis of mutual funds and local investments, and remember, this is looking at data of 1975-1994. We look at what percent of mutual fund investments over that period, based on the quarterly reports, what percent of them are held and stocks where the firm headquarters is within 100 kilometers of the mutual fund headquarters. Well, that's 6.95 percent of holdings. If the mutual fund happened to be investing in the market, 6.16 percent of their holdings would be in local stocks. That suggests there's an active tilt of about 0.8 percentage points. More money being and local firms than if the fund was just investing in the market. Not a huge difference here, but there is a little local tilt to the portfolio. Now let's cut to the chase, and get to what are the return differentials. Remember these returns, while we're looking at quarterly data, they're presented on an annualized basis with the T-statistics below them in parentheses. We see the raw excess return. Remember, we're looking at the period 1975 to 1994 where they did include not so pleasant stock markets. The local holdings, on average, have an annual return of 8.7 percent. The distant holdings have an annual return a six percentage points, therefore, this difference of local holdings outperforming distant holdings by 2.7 percent with a T-statistic over three. A statistically significant result. These are just looking at raw returns, in excess of the treasury bill rate here. When we look at this distance, it isn't accounting for risk at all. Let's account for risk. Let's look at risk-adjusted returns of local fund holdings. Once we account for risk, we see this difference in returns falls a little bit, but it's still 1.2 percentage points on an annual basis with a T-statistic greater than two. How about if we look at the difference in returns of local stocks held versus local stocks not held on an annual basis. RL here is looking at the performance of local stocks held of the mutual fund minus the performance of stocks that aren't held by the mutual fund. You see this is pretty informative. If you see a Boston company held by Boston mutual fund, that outperforms a company that isn't held by Boston mutual funds by three percentage points over the next year. There is some useful information looking at both what Boston companies or Boston mutual fund is holding, what companies are they not? Then the final return calculation here is, looking at what's the performance of local holdings that had an increase in local mutual fund ownership versus those that had a decrease in local mutual fund ownership? We're looking at just changes in how many shares are the Boston mutual funds holding in the Boston companies. If this is a firm where the share ownership of Boston mutual funds went up, it's outperforming a company where the share of Boston mutual funds went down by about 1.2 percentage points on an annual basis. Now let's think about these results, and let's think about the timing of when we're measuring these returns. The prior returns I showed you, they were actually measured starting from the quarter end. Let's think about the timing. Suppose Fidelity Magellan, using that example, they report their holdings for December 31st, 2015, on February 26th, 2016. On January 1st, 2016, only Fidelity Magellan knows what it was holding on December 31st, 2015. They're the only ones that have this information. By the time we get to April 1st, 2016, it's public information what Fidelity Magellan was holding at the end of December, because they release this information to the SEC on February 26th, 2016. Let's remember this timing. Holdings at the end of 2015, they're disclosed with, maybe, a little over a two-month lag. Once we go three months into the future, certainly, everyone should know, if they want to, what Fidelity Magellan was holding at the end of 2015. Suppose Fidelity Magellan holdings, as of December 31, 2015, predict stock returns January to March,2016. Is this a violation of market efficiency? That's Question 1 of 2. Now let's look at the second question. Again, we have this setup. Now let's use a holdings as of December 31st, 2015, and let's see, do they predict stock returns April to June of 2016? Is this a violation of market efficiency? Think about those two questions as you get back to the results of this very influential Coval and Moskowitz paper. It's interesting to think about the timing that you should measure returns as a researcher and you really have two options and Coval Moskowitz did it both ways. In this Table 1, they're looking at the performance following the report date of the holdings. In the case of Fidelity Magellan, when they're reporting holdings, December 31, 2015, they're seeing how those stocks held by Fidelity Magellan perform over January and February, and March of the next year. Now the market doesn't know what these Fidelity Magellan holdings are until a little over two months later. It's not necessarily a violation of market efficiency of like, hey, these Fidelity Magellan holdings as of December 31st, are predicting returns in January because the public doesn't have that information. I guess it might be a violation of market efficiency to say, Hey, an investor can beat the market, but the point is it isn't publicly available information. This test and Table 1 is really, do the mutual funds seem to be able to predict performance in the context of their local holdings because we're measuring returns right after their reported holding date of December 31, 2015, and we see for example the next three months after the holdings' date December 31st, March 31st, June 30th, September 30th, on an annualized basis, local holdings outperform far away holdings by about 1.2 percentage points. We also observe that local stocks held outperformed local stocks not held by about three percentage points on an annual basis. Then local stocks where you have an increase in ownership outperform those who have a decrease in ownership by 1.2 percentage points. Let's remember these differences in return across these three measures, 1.2,3,1.2 percentage points on an annual basis. In Table 2, Coval Moskowitz see what happens once this information gets released to the public. Is there still return predictability? In this case, if you can still predict returns with a three-month lag after three months, this information would be known to the public. It would suggest like, hey, that's more of a violation of market efficiency because everyone knows Fidelity Magellan holdings, not just Fidelity Magellan, what they were back in December 31st of 2015. So hey, that's publicly available information. It's more of a violation of market efficiency. This publicly available information can predict future returns, but Coval Moskowitz find precisely this. Let's have a lag of three months. Using the Fidelity Magellan example, we see what their holdings were as of December 31, 2015, but we're only going to invest in those holdings or see what the returns to those holdings are starting April 1, 2016. We're not going to start looking in January to see how those holdings do. We're going to wait three full months after they were released and look what was their performance in April, May, and June of 2016 and you see looking here across these three measures, the returns, one percent for local beating distant holdings 2.9 percent for local stocks held by mutual funds relative to stocks not held by local mutual funds. This almost one percent difference of local stocks have an increase of local mutual fund ownership versus those that have a decrease. These return differentials are almost the same that we had earlier here. That suggests that there is, the market wasn't pricing in, at least during this period, 1975, 1994, the information content of local holdings of mutual funds because you still have this return predictability even with a three month lag. But once you go out with a six-month lag, all these things, well, two of the three measures have returns of zero. There still seems to be some predictability looking at the difference of local stocks held relative to non-local stocks held. That was the bigger lag but you see the return differential start to diminish the longer you go into the future. To dig deeper, let's just not look at all mutual funds, local and non-local holdings, and what local stocks are investing and what local stocks that are not investing in. We can actually think of breaking mutual funds into 20 percent bins where Group 4 and Group 5 represent those mutual funds in the top 40 percent of the distribution in terms of their local bias. Particularly Group 5, is the 20 percent of funds with the biggest local buyers. You can see that about 22.5 percent of their holdings are in local stocks. If they were just investing in the market, this would have been about 8.7. Group 4 as well also has a positive local bias, and you see when you look at the numbers here in terms of how to local stocks perform relative to far away stocks? How do local stocks perform relative to local firms not held, you see bigger differentials once you focus on mutual funds that have a bigger local bias. If you're implementing, this is a strategy to try and mimic, you would say, hey, let me know the local stocks Boston funds are investing and let me know the stocks they're are not investing in and let me even focus deeper on Boston mutual funds that have a big local bias because that might be a signal that they have even more information behind their stock picks, their putting their money where their mouth is. Here you see for these groups, that's where to really focus. Now, let's think about sorting portfolios based on the local ownership of firms. When we're looking at the local ownership of firms and we're putting them in groups, Quartile 1 versus Quartile 5 here, and then looking for differences. How do we think about doing this? Well, let's look at for a given company, let's see what's the percent of mutual fund investment in that company that comes from local funds. Let's then compare that to what is the fraction of all mutual fund dollars that come from local funds. It's like a relative local ownership, and this is a way to treat New York firms differently than Wyoming firms like New York firms might have a lot of local ownership just because a lot of funds are located in New York. You want to know, do New York firms have relatively higher local ownership by doing a benchmark of mutual funds investing in the New York company, what fraction are from New York and then just compare that to what fraction of all mutual fund dollars are from New York. That's how this local ownership is a calculated for all the firms. Let's look at the results, probably not surprising at this point. Those firms that have higher local ownership, more local ownership by local mutual funds outperform those with less ownership by 2.6 percent on an annual basis and that t-statistic is greater than two, so it says statistically significant result. Once you control for the Alpha risk-adjusted in a four-factor model, this difference declines, but it's still 1.1 percentage points and statistically significant here at the 10 percent level, T stats just below two. Coval and Moskowitz also break results by firm size so looking at this, focusing on small firms where we think there should be the largest effect. That's exactly what you find. You find small firms that have a lot of investment from local mutual funds. Their future performance is almost 18 percent per year. Those that has small local ownership, it's only 13.5. There is difference of four percentage points. Once you risk adjust, you see a tape, basically the same thing. It goes from 4.2 percent difference in raw returns to 3.6 per cent difference and alphas once you do the four-factor models. Definitely amongst small firms, if you see local mutual funds investing in them over this period, 1975-1994, that was a good sign about future things happening for the firm. But when you look at large companies, whether a local mutual fund is investing in them or not, makes absolutely no difference in the future return, whether you look at raw returns or whether you do a risk adjustment. Local mutual fund ownership only matters for predicting the returns of small firms. This mirrors what we found for individual investors in Module 3. Further work on the geography of investment building on this seminal piece by Coval and Moskowitz, Baik, Kang and Kim also study the role and performance are geographically close institutional investors. What role they play in the stock market and pricing. They focus on all institutional investors as opposed to just mutual funds. All institutional investors at disclose quarterly holdings with the Securities and Exchange Commission. Institutions managing more than $100 million and US equities have to file a quarterly report other holdings with the SEC. So then Baik, Kang and Kim can just look at all these holdings, see if there are local investment or not. What do they find? Well, they find results similar to Coval and Moskowitz with a more updated dataset, they find that both the level and the change of local institutional ownership in a firm's stock predict positive movements in that firm's stock going forward. We always want to look beyond the horizon and experience new things. But when it comes to individuals and mutual funds, it seems historically that lot size, if you want to see which investments perform the best, it's their investments that are in their backyard.
