Go with the flow
A share portfolio mimicking stock sales in different market sectors could outperform an S&P tracker by 40%, research suggests. Steve Johnson reports.
"It's puzzling. You shouldn't be able to make money that easily just by ranking stocks in order." Building an investment portfolio that can consistently beat a leading stock market index by 40% - with superior risk and return calculations thrown in for good measure - is the Holy Grail for investors. But it could be relatively straightforward to create, if three academics are correct. Alessandro Beber, Professor of Finance at Cass, Michael Brandt, Professor of Finance at the Fuqua School of Business at Duke University in North Carolina and Kenneth Kavajecz, Professor of Finance at the Wisconsin School of Business, reveal their secrets in a paper entitled What Does Equity Sector Orderflow Tell Us About the Economy? As the title suggests, the trio concentrated their research on orderflow data, the collective buy and sell orders sent by brokers to dealers, to determine whether it could predict the future strength of the economy, and thus the performance of equity and bond markets. Their number crunching suggested it had "striking economic implications". They found that a portfolio that mimicked orderflow, constructed by over or underweighting market sectors depending on the signals from orderflow data, would have turned $100 invested in the US stock market in 1993 into $350 by 2005, comfortably outstripping the $250 that would have been achieved by a portfolio passively following the S&P 500 index.
Moreover, the orderflow-based portfolio, measured against the market, would have exhibited a lower standard deviation (a measure of the risk associated with price fluctuations) and a significantly higher Sharpe ratio (a measure of risk-adjusted returns from an investment strategy - the higher the Sharpe ratio, the higher the returns). Further, the portfolio would be relatively defensive, underweighting cyclical sectors such as information technology, materials and industrials ahead of downturns in the economy. "It's clear that this approach would have reduced your IT exposure before the dotcom crash in 2000-01," says Professor Beber. The fact that such a portfolio can beat the market might not come as a surprise to some, even though the extent of the outperformance can be eyebrow-raising. The orderflow approach has some similarities with momentum trading, which involves buying a basket of stocks that have risen in value in the previous time period, and taking short positions on stocks that have fallen in value. Research has shown that this, too, tends to produce market-beating returns, although it remains a mystery why such a simplistic strategy should succeed in contravention of the efficient-market hypothesis, which asserts that consistently beating the market given the information available at the time of the investment is not possible. Paul Marsh, emeritus Professor of Finance at London Business School, who has studied the phenomenon, admits: "It's puzzling. You shouldn't be able to make money that easily just by ranking stocks in order."
Predicting price changes
However, Professors Beber, Brandt and Kavajecz appear to demonstrate that their approach is better than momentum strategies. One might expect changes in orderflow to result in identical changes in stock prices, given that the level of trade in a market sector determines share prices. However, the research suggests that this mechanism is less than perfect, with an increase in net demand for a given sector not necessarily producing a similar rise in prices. Crucially, the trio's work suggests that orderflow data carries more "information" than price fluctuations. That is, it is better able to predict future economic and market movements. For instance, by using orderflow data it is possible to forecast the level of an economic indicator three months in advance twice as well as by extrapolating from the current level of the indicator. The study established this by applying the data to the Federal Reserve Bank of Chicago's National Activity Index. In contrast, analysing the sector returns improved forecasting ability by only 2%, the paper found. "Most theories would assume that whatever we see in the orderflow shows up immediately in the returns. In reality I think it's very different," says Professor Beber. "It's very difficult to pin down the specific reasons why flows are much more informative, but it's clear from our data that they are. Therefore there must be some sort of friction that prevents all the information that is in the orders showing up in the returns." Using orderflow also allows the observer to concentrate purely on larger buy and sell orders. The team found that focusing on trades of more than $250,000 - likely to be placed by more sophisticated institutional investors - improved the forecasting power of the data.
So are any investors already using such a trading strategy? Investment banks with strong market-making arms would already have access to their own slice of orderflow data and, ahead of the implementation of the Volcker Rule in the US, still maintain proprietary trading desks that would be well placed to take advantage of this information. Professor Beber believes some banks probably have put two and two together. "I know that investment banks are looking at flows to identify particular trading patterns," he says. The New York Stock Exchange makes orderflow information, in the shape of its so-called consolidated tape, available publicly for fees ranging from $1 a month for private investors to $2,000 a month for computer-based quantitative trading houses that plug the entire data feed into their systems to search for patterns. Its main electronic tape has 2.84 million subscribers. However it is possible that the power of the orderflow book might be waning, even as more investors wake up to its forecasting power. The emergence of off-exchange trading venues, such as "dark pools", allows institutional investors more scope to hide their trades and also means the overall data flow is more fragmented. Also, more institutions are using algorithms to split trades into a series of smaller deals, meaning they would not show up in datasets that rely on identifying trades by larger market participants. Professor Beber admits that the ability to spot institutional trades from large orders was stronger in the first half of the study than in the second. However, given the strong outperformance identified by the study, a modest weakening of the model's predictive powers should not prove fatal. In a world where traders scramble to profit from the tiniest possible arbitrage opportunity, potential double-digit returns are not to be sniffed at.
Steve Johnson is Deputy Editor of FTfm at the Financial Times.