Quantitative Stock Selection: Transfer Coefficients

1
Quantitative Stock SelectionTransfer
Coefficients

• Campbell R. Harvey
• Duke University
• National Bureau of Economic Research

2
Transfer Coefficients

• Certain factors induce excessive turnover.
• Example is a reversal factor, such as a lagged
  return.
• Selection routine needs to balance the
  information in the signal with the cost of
  implementing

3
Transfer Coefficients

• 4. One approach is to limit turnover. For
  example, a factor might induce 120 annual
  turnover. One might impose a constraint to limit
  the turnover to approximate 60 per year.
• 5. To implement in terms of scoring screens,
  suppose in any month about 10 of the stocks
  leave or are added (120 per year).
• (a) Calculate the difference between the old
  scores and the new scores. Consider the highest
  25th percentile moves and the lowest 25th
  percentile moves. Ignore the middle.

4
Transfer Coefficients

• 5. (b) Consider the old ranking and the new
  ranking. Only consider the stocks that move up
  and down a lot. For example, if you have
  quintiles with 100 stocks in each, and stock XYZ
  drops from rank 100 (Q1) to 101 (Q2) you do not
  sell (assuming you are long Q1 and short Q5)

5
Transfer Coefficients

• 6. Alternatively, one can smooth the signal. For
  example, instead of looking at RR you look at a
  three month moving average of RR. The smoother
  factor will induce less rebalancing.
• 7. Both of these techniques degrade the signal
  from the factor. The transfer coefficient is the
  correlation between your signal changes and your
  weight changes.