Chapter 5: China A-Share Mutual Funds

This is the fifth installment in a series of articles on Investing in China A-Shares. The prior articles can be found here:

1)     Allocating to China A-Shares

2)     Overview of China A-Shares

3)     State-Owned Enterprises

4)     Regulatory Releases

Quick and important disclaimer: The analysis and views expressed in this discussion are my own and do not necessarily represent the views of my employer or any organization I am affiliated with. Nothing in this article constitutes investment advice or a recommendation to buy or sell any security.

Mutual funds in China A-shares outperform the market even after fees. Their performance is persistent. In particular, mutual funds with high return and information ratios in one year tend to have return and high information ratios the next month, next 3 months, next 6 months, and next year as a whole. The excess returns of mutual funds tend to come from stock selection as opposed to market timing. Despite the fact that mutual funds only release complete holdings twice a year and at a delay, matching their stock holdings can earn you alpha but just barely.

But mutual funds outperforming is important for reasons beyond using their holdings as a signal. While you cannot invest in onshore mutual funds without a qualified foreign institutional investor (QFII) quota, the outperformance of mutual funds tells us about the inefficiency in the A-shares market. If mutual funds outperform and that performance is persistent, then that suggests that the market is inefficient in a way that can be persistently captured. Whatever the skilled managers are doing will continue to work in the future. Contrast this to the US market where 90% of large cap managers underperform their benchmark over any 10-year period and there is no persistence in return. That is a market where it is exceedingly difficult to earn alpha. It is likely better to hold an index fund for the US market, while it may be sensible to implement an active strategy in China A-shares.

Manager Outperformance

The first thing to note, as we did in the first chapter, is that mutual funds outperform on average. We replicate the graph we showed in Chapter 1 below with one additional line—the capitalization-weighted market as a whole in blue.

Most managers choose the CSI 300 as a benchmark and yet their returns more closely match that of the broad market. The CSI Equity Fund’s correlation with the CSI 300 is 0.91, but its correlation with the full China A market is 0.96. Since the average China A fund holds a large cash reserve, the beta of the CSI Equity Fund Index is 0.85 with respect to the CSI 300 and 0.87 with respect to the market. Assuming that they hold on average beta 1 stocks, that suggests they hold about 13% cash or bonds at any given time. Holdings-based analysis yields similar results. Those cash and bond holdings are massive relative to US equity funds, which generally try to stay close to fully invested.

Because the fund index has a low beta and outperforms, naturally, the alpha will be positive if the market return is positive. Indeed, the CSI Equity Fund Index earns a weakly significant (threshold p-value 0.10) CAPM alpha of 4.7% against the CSI 300 and 3.5% against the full market from December 2007 to December 2020. What makes this performance more impressive in some ways is that these funds perform this feat after fees and the median management fee for a Chinese equity fund is 1.5%. That excludes service fees and transaction costs, which are high given the high turnover of these funds.

The next question is where does this alpha come from? If it comes from stock selection, we have some hope that we can use the holdings to build a cross-sectional signal. If it comes from market timing, then there is no use in looking at holdings except insofar as funds hold cash versus stock. While we will replicate these results using holdings, the best way to measure this using only returns is to measure monthly alphas and compare that to using full period alphas. Full period alphas include a market timing component but running a regression within each month will generate the alpha relative to the CSI Equity Fund Index’s specific beta in that month. The average monthly alpha across all months reflects the stock selection alpha. The difference between the full period alpha and the stock selection alpha is the market timing alpha.

When viewed against the CSI 300 Index, managers earn a -2.3% alpha from market timing but more than make up for that with a 7.0% alpha from stock selection. But we have already discussed why the CSI 300 index may not be the best mutual fund benchmark. When viewed against the full China A-shares market, managers earn 83 bps from market timing and 2.6% from selection. That is a bit more modest, but we can conclude managers earn the vast majority of their alpha from stock selection and that we may be able to earn excess returns by using stock holdings as a signal. Before we look into that, we will discuss performance persistence because it will play into our conversation about holdings.

Performance Persistence

Now, we know that fund managers as a whole outperform. Before we go further, we ought to ask ourselves whether we expect mutual fund performance to be persistent. Fund managers demonstrate persistent skill as a whole, so it is reasonable to expect that there would be a cross-section of skill among fund managers. Indeed, it would be strange to imagine that fund managers were more skilled than the rest of the market but identically skilled to one another such that any performance differential among them is luck.

For all further analysis, I only have data ending in 2019 for individual funds. First, we will look at returns. We will sort funds on returns in the prior year into quintiles and determine how they perform in the subsequent year. If funds have strong returns in one year, will they have strong returns the next? Yes.

However, it turns out that this is a profoundly misleading graph. Different funds hold different amounts of cash and do so persistently. That means that funds that own more stock will mechanically outperform persistently since the market has gone up over time. What about alpha? Do funds that earn high alpha in one year tend to earn high alpha the next? No, as we can see from the graph below.

But this is misleading in another way. It does not account for benchmark risk. Chinese mutual funds tend to be extremely concentrated so those that earn very high or very low alpha are often taking massive risk and are just getting lucky or unlucky, respectively. What about information ratio or modified information ratio? Those are indeed both persistent. For brevity, we will only show modified information ratio here, since alpha scaled by the standard deviation of the CAPM residual best captures what we are trying to measure. Modified information ratio is just the beta-adjusted information ratio and is sensible for evaluating strategies with beta different from 1.

Modified information ratio is fairly stable especially for the highest IR funds. But modified information ratio can be high for two reasons—managers are good at generating alpha or they are good at managing risk. Let’s try to use modified IR to predict alpha.

It works decently well. The high minus low portfolio earns a significant 2.1% excess return (t-stat: 2.77). However, it is not even close to monotonic. Moreover, given the result is weaker than the prior graph, it suggests that managers earn high modified IRs largely through skillful risk management.

Now, we will look at holdings, but since modified IR only weakly relates to future alpha, we will find that using the holdings from the highest modified IR funds only performs slightly better than using holdings from all funds. It is critical to understand that information ratio does not average across funds the same way alpha or return might. The reason why high information ratio funds have high information ratios is because they earn positive excess returns and engage in strong risk management. The reason why low information ratio but positive return funds have low information ratios is that they fail to adequately manage risk, but when you diversify across low information ratio funds, you effectively create a risk managed portfolio out of a set of poorly risk-managed funds. If they are decent at earning alpha but poor at risk management, their poor risk management will wash out when aggregated. Of course, when we use holdings of all funds in a group, we are aggregating, so the benefits of using the highest information ratio fund will be weaker.

Information in Holdings

That is not the only issue in capturing information from fund holdings. Even if you use all fund holdings, we run into two issues. Funds release their full holdings twice a year and do so at a roughly 2-month delay. The second related issue is that funds have an average one-way turnover of 340% or two-way turnover of over 680%. (See Yi, Liu, He, Qin, and Gan's 2018 paper "Do Chinese mutual funds time the market?") Even if you had their holdings without a delay, they would become utterly stale in months. With these two headwinds, there is only minimal information in holdings. If you buy the full January and July holdings of funds at the beginning of April and October, respectively, you would earn an insignificant 2.70% annual CAPM alpha (t-stat: 1.27) at a 0.98 beta.

If you buy just the holdings from the highest information ratio funds, you would earn a 3.40% annual alpha (t-stat: 1.57) at a 0.98 beta. That reaches a p-value of 0.11 but is hardly inspiring. Indeed, it is a far cry from the alphas of the funds themselves despite the fact that funds charge massive fees. Again, this is due to delayed holdings and massive turnover of the funds.

What about market timing information? If you bought the market in proportion to the market holdings of funds, you would earn an annual CAPM alpha of negative 80 bps (t-stat: -1.00). Funds demonstrated no skill at market timing when we looked at returns. It stands to reason that they would demonstrate no skill when we look at holdings, especially at a 3-month delay.

Chinese mutual funds achieve strong alphas despite exorbitant fees. We had hoped that we could earn those returns without the fees by using their holdings. Unfortunately, their holdings come out too infrequently at too long of a delay and the funds have too high turnover for the holdings to have anything other than the most marginal of information. While we cannot use their holdings to earn alpha, their performance shows it is possible to earn alpha, which suggests we are on the right track. China A-shares is a place where alpha can be found. Now, I will show you how you can earn that alpha.