Shedding Light on “Invisible” Costs: Trading Costs and Mutual Fund Performance
Shedding Light on “Invisible” Costs: Trading Costs and Mutual Fund Performance
复制标题
揭示“隐形”成本：交易成本和共同基金业绩
DOI：
10.2469/faj.v69.n1.6
复制
发表时间：
2013
影响因子：
2.8
通讯作者：
Gregory Kadlec
中科院分区：
文献类型：
--
作者：
Roger M. Edelen;R. Evans;Gregory Kadlec
行业观察人士长期以来一直对基金交易的“无形”成本提出警告，但证明这些成本重要性的证据好坏参半，因为许多研究没有考虑到最大的交易成本成分价格影响。通过使用投资组合持有和交易数据，作者发现，基金的年度交易成本平均高于其费用率，并对业绩产生负面影响。他们还开发了一个准确但计算简单的交易成本代理头寸调整的营业额。费用比率是共同基金回报表现的少数可靠预测因素之一，低成本指数和交易所交易基金的市场份额不断增加，表明投资者在做出投资决策时会使用这些信息。然而，正如约翰·贝恩斯和其他著名的行业观察家所指出的那样，费用比率只反映了“可见”（即，（二）共同基金的费用。基金产生了大量对投资者不太透明的“无形”成本最明显的是，与投资组合头寸变化相关的交易成本。在我们的研究中，我们估计了基金的年度交易成本支出，并研究了这些成本对基金回报表现的影响。我们利用基金投资组合持有数据、交易级证券数据和美国证券交易委员会备案文件，开发了一个详细的按头寸衡量基金年度交易成本支出的方法。首先，我们使用季度投资组合持有数据来确定每只基金的头寸变化。第二，对于每一个头寸变动，我们都估算了该季度交易该股票的成本（经纪佣金、买卖价差和价格影响）。第三，我们计算了每只基金的年度交易成本支出，方法是将该基金一年中所有交易的成本相加。我们将这种方法应用于1995-2006年的1,758只国内股票基金样本。我们发现，基金的年度交易成本支出（即，总交易成本）与费用比率（分别为1.44%和1.19%）的幅度相当。此外，基金交易成本的变化远大于费用比率的变化。例如，小盘成长型基金和大盘价值型基金的平均费用率差异为0.32个百分点（1.39%对1.07%），而同一基金的平均总交易成本差异为2.33个百分点（3.17%对0.84%）。更重要的问题是基金在交易成本上的支出与回报表现之间的关系。我们发现，总交易成本和基金收益表现之间存在很强的负相关关系。按费用、基金总净资产或营业额（最常见的交易成本代理）对基金进行排序，不会产生一致、单调的回报模式。与此形成鲜明对比的是，根据总交易成本估算对基金进行排序，得出了一个明显的单调模式，即随着基金交易成本的增加，风险调整后的业绩会下降。总交易成本最高和最低五分之一的基金平均年回报率之差为-1.78个百分点。鉴于总交易成本在预测基金业绩方面的力量，它将是投资决策者的有用工具。不幸的是，由于数据可用性和计算复杂性的原因，这些对基金交易成本的直接估计很难获得。学术界和从业者都使用的最容易获得的衡量交易成本的指标是基金周转率。然而，关于基金周转率与收益绩效之间关系的实证研究却并不明确。我们确认，这种模糊性是由于成交量不考虑基金交易的差异成本-这取决于基金规模（即，交易规模）和股票流动性（即，小盘股对大盘股）。例如，一个5亿美元的小盘基金，50%的营业额将比一个1亿美元的大盘股基金，100%的营业额高得多，尽管前者的营业额较低。为了解决这一潜在缺陷，我们建议对营业额进行简单调整。特别是，我们通过将每个基金的成交量乘以其相对头寸规模来计算“头寸调整成交量”。基金的相对持仓规模等于其平均持仓规模（总净资产除以持股数量）除以其市值类别中所有基金的平均持仓规模。相对头寸规模反映了基金交易的价格影响，这是基金交易成本的最大组成部分。我们发现，这个简化的代理具有类似于我们更挑剔的措施的权力。仓位调整后换手率最高和最低五分位的基金，其年均回报率之差为-1.92个百分点。总的来说，我们的研究结果表明，交易成本是基金业绩的一个重要决定因素，我们提供了一个简单的交易成本代理，可供投资者和研究人员使用。
Industry observers have long warned of the “invisible” costs of fund trading, yet evidence that these costs matter is mixed because many studies do not account for the largest trading-cost component—price impact. Using portfolio holdings and transaction data, the authors found that funds’ annual trading costs are, on average, higher than their expense ratio and negatively affect performance. They also developed an accurate but computationally simple trading-cost proxy—position-adjusted turnover. The expense ratio is one of the few reliable predictors of mutual fund return performance, and the increasing market share of low-cost index and exchange-traded funds suggests that investors use this information when making investment decisions. However, as noted by John Bogle and other prominent industry observers, the expense ratio captures only the “visible” (i.e., reported) costs of mutual funds. Funds incur a host of “invisible” costs that are less transparent to investors—most notably, the transaction costs associated with implementing changes in portfolio positions. In our study, we estimated funds’ annual expenditures on trading costs and examined the impact of those costs on fund return performance. We developed a detailed position-by-position measure of funds’ annual expenditures on trading costs by using fund portfolio holdings data, transaction-level securities data, and U.S. SEC filings. First, we used quarterly portfolio holdings data to determine each fund’s position changes on a stock-by-stock basis. Second, for each position change, we applied an estimate of the cost (brokerage commission, bid–ask spread, and price impact) of trading that amount of that stock in that quarter. Third, we computed each fund’s annual expenditure on trading costs by aggregating the costs of all trades for that fund over the year. We applied this approach to our sample of 1,758 domestic equity funds over 1995–2006. We found that funds’ annual expenditures on trading costs (i.e., aggregate trading cost) were comparable in magnitude to the expense ratio (1.44% a year versus 1.19%, respectively). Moreover, there was considerably more variation in fund trading costs than in expense ratios. For example, the difference in average expense ratio for small-cap growth and large-cap value funds was 0.32 percentage points (1.39% versus 1.07%), whereas the difference in average aggregate trading costs for the same funds was 2.33 percentage points (3.17% versus 0.84%). The more important question concerns how funds’ expenditures on trading costs relate to return performance. We found a strong negative relation between aggregate trading cost and fund return performance. Sorting funds by expenses, fund total net assets, or turnover (the most common trading-cost proxy) yielded no consistent, monotonic pattern of returns. In stark contrast, sorting funds on the basis of their aggregate trading-cost estimate yielded a clear monotonic pattern of decreasing risk-adjusted performance as fund trading costs increase. The difference in average annual return for funds in the highest and lowest quintiles of aggregate trading cost was –1.78 percentage points. Given the power of aggregate trading cost in predicting fund performance, it would be a useful tool for investment decision makers. Unfortunately, these direct estimates of fund trading costs are difficult to come by for reasons of both data availability and computational complexity. The most readily available metric to proxy for trading costs, used by both academics and practitioners, is fund turnover. However, the empirical evidence on the relation between fund turnover and return performance is ambiguous. We conjectured that this ambiguity is due to the fact that turnover does not account for the differential cost of fund trades—which depends on fund size (i.e., trade size) and stock liquidity (i.e., small cap versus large cap). For example, a $500 million small-cap fund with 50% turnover will have much higher trading costs than a $100 million large-cap fund with 100% turnover, despite the former’s lower turnover. To address this underlying deficiency, we propose a simple adjustment to turnover. In particular, we compute “position-adjusted turnover” by multiplying each fund’s turnover by its relative position size. A fund’s relative position size is equal to its average position size (total net assets divided by number of holdings) divided by the average position size of all funds in its market-cap category. Relative position size captures the price impact of the fund’s trades—the greatest component of a fund’s trading costs. We found that this simplified proxy has power similar to that of our more fastidious measure. The difference in average annual return for funds in the highest and lowest quintiles of position-adjusted turnover was –1.92 percentage points. Overall, our results suggest that trading costs are an important determinant of fund performance, and we offer a simple proxy for trading costs that can be used by investors and researchers alike.