skip to main content
Close Icon We use cookies to improve your website experience.  To learn about our use of cookies and how you can manage your cookie settings, please see our Cookie Policy.  By continuing to use the website, you consent to our use of cookies.
Global Search Configuration

Stock flows: Building the best mouse trap

EPFR has a long history of helping clients understand where money is moving and how fund managers are investing that money. These insights started at the regional, country and sector level. In recent years, this coverage has been extended to the industry and single security level.

The stock-flow offering, based on monthly stock-level positioning of mutual and exchange-traded funds, provides a new kind of fuel to drive alpha-generating models. But formulating that fuel correctly and applying the right filters takes time.

Several years have passed since we launched our stock flows product, and we have used that time to analyze, test and experiment with the data to develop and improve our own multi-factor flow alpha model.

What have we learned?

Factoring in peak performance

The conditions under which EPFR collects the stock-level positioning data preclude us from passing it directly on to clients. We can however report aggregations of, or quantitative factors computed on, the raw data.

In the process of digging into the stock flow dataset, EPFR has developed many quantitative factors that are now part of our standard offering. Some of these have been combined into the alpha model which, after several years of revision, adjustment and back testing, optimizes the forward-return potential of this model.

The factors – out of some two dozen-- that have earned a place in this model are:

  • Active Weight Trend – a measure of the extent to which good managers, the ones receiving inflows, are overweight a stock with respect to their peers. Formally, this is ΣφΔω⁄Σ|φΔω| where Δω is active weight versus peers and φ is the daily flow into each fund. This factor is averaged over the last three weeks before being incorporated in the model.
  • Active Weight Diffusion – a parameterized version of Active Weight Trend, the active weight having been replaced by its sign. This factor is also averaged over the last three weeks before being incorporated in the model.
  • Flow Diffusion – a measure of the extent to which good managers, the ones receiving inflows, are raising allocations to each stock. This factor is the same as Active Weight Diffusion, except that we are using the month-over-month change in weight of each fund rather than that fund’s active weight with respect to its peers. This factor is averaged over the last eight weeks before being incorporated in the model.
  • Allocation Trend – a measure of the proportion of managers increasing weight to each security. Formally, this is ΣαΔω⁄Σ|αΔω| where Δω is the change in weight over a month and α is the average of starting and ending assets. This factor is averaged over the past eleven months prior to incorporation into the model.
  • Allocation Skew – the skew factor of Srimurthy et al. (2018)[3]) applied to stock-level holdings rather than country weights of funds. This is the asset-weighted percentage of managers underweight a security with respect to their peers. This factor is averaged over the past eleven months prior to incorporation into the model.
  • Herfindahl – the Herfindahl dispersion of the fund holdings in each security, the factor introduced in Barabanov (2003)[1].
  • Fund Count – count of funds holding a security, the factor introduced by Chen, Hong and Stein (2002)[2].


Each factor has a 10% weight in the model, except for Allocation Trend and Fund Count. The weightings on these factors are 20% and 30% percent respectively.

What do they add up to?

We use this flow alpha model to identify the countries in which EPFR stock-flows data (a) has historically proved most effective and (b) measure the degree to which this is true.

The countries investigated are those which had a minimum of 10 companies listed in the MSCI World or MSCI EM indices at the beginning of each month from July 2011 to January 2020.

For each country we rank monthly, based on flow alpha, the stocks from that country in the MSCI World or MSCI EM indices into five buckets. Because we only look at countries with 10 or more companies listed in the indices, we are guaranteed a minimum of two observations in each bucket.

These quintile portfolios are held for a month and then rebalanced.

The table below shows two things: the annualized average quintile spread -- the return to the highest flow alpha bucket in excess of that to the lowest -- and the associated annualized Sharpe ratio.

Quants Corner


Based on the results, EPFR’s stock-flows offering works better in European developed markets than Asian or North American ones. Among emerging markets, it does well in the Asian Tiger economies.

Although currently refined to a satisfactory degree, the model can and will be tweaked as new factors developed either by EPFR or its clients become available. 

For more insight subscribe here.

Any questions? Speak to a specialist

Would you like to request sample data or analysis from Informa Financial Intelligence? 

See how our tailored solutions can help you gain a competitive advantage: