Reputation and Risk-Taking By Mutual Fund Managers

In the last post (paste link here) I discussed how the reputation of a mutual fund plays a crucial role in determining how investors react to fund's recent performance. In this post, I shift my attention from investors to mutual fund managers. Here I discuss some results from my latest work on the behavior of fund managers. By behavior I mean the way manager forms his portfolio (portfolio is nothing but a bundle of asset and it is usual in finance it sometimes also means the investment strategy) and manage the risk in his investments. In my paper I study how the historic performance of a fund alters its risk taking behavior. Before proceeding, let's understand what matters for fund manager. Managerial incentives are two dimensional: on one hand he has to attract capital from investors as ultimately his compensation depends upon the size of the fund he manages. On the other hand he is answerable to fund management and his employment contract depends upon his performance. We have already seen in the last post that investor reaction depends upon historic performance together with current performance. Data suggest that reputation or history of performance also matters for employment continuation. In my sample, 30\% of the managers who are fired from their job ranked within bottom 20\% in terms of their historic performance. Clearly, having a poor history raises the probability of firing. (Else the number should have been 20\% instead of 30\%. Think!) Clearly we know now that a poor history fund manager is facing a serious unemployment risk as compared to a good history fund manager. So there is a strong case to believe that reputation must matter for the way managers handle their investments during current period.

In 1996, Brown, Harlow and Starks (BHS hereafter) came up with a study which found that a midyear losing manager is more likely to bump-up the risk of his portfolio in the second half of the year. But later studies did not find strong support for this evidence. Kemph, Ruenzi and Theile published a research in 2007 doing a very simple trick. They argued that compensation incentives are more dominant during the period when stock market performs well but employment incentives are strong during bad years. So managers might manage risk differently over good and bad years. I combine Kemph's approach and include the reputation as a new driver of managerial risk taking ability. So we are interested in analyzing how managers change their investment risk profile in the later half of the year as compared to first half of the year, given their mid-year performance position. It is important to know a good measure of risk in this context: it is precisely the extent to which manager deviate away from what their peers are doing. So risk is measured using how much manager's returns deviate away from benchmark. Then change in risk is just risk in the second half of the year minus risk in the first of the year. I prefer to take the ratio of second half risk to the first half risk, so that the ratio of 1 is status quo.

What I find is pretty interesting. First I find that midyear position has no impact whatsoever on risk-shifting during good years. Only during bad years, when unemployment risk is higher do the managers try to adjust the risk. Following figure shows the results for bad years: On X-axis we have mid-year rank (0 being lowest and 1 being highest) and on Y-axis we have risk-shifting ratio: ratio of risk during second half to first half. What we have on our hands is really a fascinating result: As the mid-year performance gets better, top reputed funds reduce this ratio and low reputed funds increase this ratio. (Note the ratio is more than 1 for any reputation and mid-year position. So it is a general trend that all funds bump-up their risk during second half anyway. We want to understand how the extent of that change with mid-year performance).  This is a dramatically opposite behavior.

How can we explain it? There are two dimensions we need to explain. First is how the manager behavior change as mid-year position change and second comparing the levels of the risk-shifting for top and low reputation managers. I answer these one by one. First, top reputed manager has lower risk of getting fired with one-off bad performance, so these managers bump-up the risk more during second half of the year as compared to low reputed managers when in a mid-year losing position. That explains why the curve for top guys starts at a higher level. Second, as mid-year performance improve, top guys want to play it safe. This is because even a reasonable performance during the year assures them employment next period and also we know from my last post that a bad performance by top guys is penalized more by investors. So given a reasonable mid-year position there is no point in bumping up the risk. That explains why risk-shifting curve slopes downwards for top guys. Now coming to low reputed managers: For low reputation guys, if they are in a reasonable mid-year position, it gives them some head-start to capitalize and they have incentive to bump-up the risk. This is gambling for resurrection. But as their mid-year performance gets bad, their risk appetite goes down very much as unemployment risk bites them. So their objective is not to be a topper anymore but just be a median guy, so they keep if very safe, having portfolios very similar to their peers. This explains why curve slopes upwards for low reputed guys.

Of course all these are conjectures and it is important to write down a proper theory to see if this logic holds in a rational model. But what the results show is very interesting and important from the compensation designing perspective. Fund management teams can optimally construct the salary schemes so as to keep managerial risk-shifting in control and what my results shows is that managers need to have element in their salary to hedge against unemployment risk. Otherwise, managers are probe to risk-shifting!

Mutual Fund Reputation and Investor's Reaction !! (My PhD Job Market Paper)

Does mutual fund's historic performance matter while investors determine their portfolios after observing a recent period's fund performance? In my PhD thesis, I solve this interesting question using a large dataset of US equity mutual funds. Before starting with the paper, I had this intuitive notion that a good history mutual fund can escape a one-off bad performance since investors are more likely to attribute this to a bad luck rather than lack of skill. Such behavior is also supported by behavioral traits such as confirmation bias where investors tend to disregard the new information if it does not match with their prior information. So a bad performance by a good history fund is exactly a type of information they are ready to ignore. But what I found in the data was exactly opposite to this intuition: A bad performance by a good-history fund experiences more fraction of capital outflow as compared to a bad-history fund. This left me with a small puzzle on my hands. Just to state findings complete, I also find that a good performance by a good-history fund attracts a lot more percentage inflows as compared to a bad-history fund. In summary, good-history funds experience very sensitive capital flows to their recent performance but bad-history funds neither lose lot of money on bad performance nor gain any significant capital with a reasonable performance. Just to give sense of magnitude, I report the numbers from my regression analysis. Consider a worst fund and a best fund. Worst fund has a bad history and also performs badly this period. On the other hand a best fund has an excellent history and it also performs nicely this period. Then best fund receives 50\% (as a percentage of asset size) as compared to worst fund. Out of which one-fifth can be attributed to the fact that best fund has a better recent period performance, one-tenth to the fact that it starts the period with higher reputation due to better historic performance, but whole of the remaining two-third to the fact that it performed well this period and it also had a good reputation to start with. This last effect is the joint effect and shows that the importance of the recent period's performance grows with good historic performance.

So there are two immediate questions: First is why this result is interesting? Second how one can explain this counter-intuitive evidence? To answer first question, it's important to know what was known till the day about capital flows in and out of a mutual fund. The notion of 'return chasing' was pervasive: meaning that investors chase recent winner funds and drop out of recent losing funds. But what my data shows is that the extent of this tendency is virtually determined by the historic performance. The good-history funds experiences this return chasing type of investor's behavior, but flows in and out of a bad-history fund are insensitive. So my results are important to qualify the applicability of this return chasing notion. Second important reason is that the investor's reaction determines the risk that a fund manager is ready to undertake. To this extent, good-history and bad-history funds can showcase very different risk taking appetite. I present the evidence on risk taking in a second post.

Coming to the second question, I have a simple story to explain these results. Imagine a world with two types of investors: Some investors are more attentive to the information than others. In this world, attentive investors always update their beliefs about the fund manager after each performance and shift out if fund keeps on performing badly. Necessarily they shift to good performing funds. Only inattentive investors stick with badly performing funds. So good-history funds are owned by attentive investors and bad-history funds are owned by inattentive investors in large part. This implies that a bad-performance by a good-history fund will be penalized more.

In summary, these results give a very different picture than what was thought to be true earlier and importantly can help funds managers understand the type of investor behavior they are likely to face given their historic performance.