1. Transactions Costs

2. Important Topic
Average manager underperforms by average amount of trading costs and fees.
Portfolio management trades off expected returns against costs incurred with certainty.
Traditionally, the importance of transactions costs has far exceeded the amount of attention paid.
The death of a thousand cuts.

3. Explicit and Implicit Costs
Transactions costs have explicit and implicit components:
Explicit components:
Commissions
Taxes
Implicit components:
Spreads, market impact, opportunity costs
These show up in the prices. We don’t separately pay them.

4. Defining and Measuring Costs
Implementation Shortfall Approach.
Treynor and Perold
Good overall framework for understanding components of costs.
Basic idea:
Compare real portfolio to a paper portfolio. Both portfolios start with same value. Real portfolio trades over interval, while paper portfolio does not. Compare final values.
Implementation shortfall is difference in value between the two portfolios.

5. Implementation Shortfall
Note: we will find it useful to work with numbers of shares, rather than our usual fractional portfolio holdings.
Paper Portfolio
ni, pi(0), pi(T)
Real Portfolio
mi(tj), pi(tj)
Trades occur at times {tj} between 0 and T.
Prices are net of commissions and taxes.

6. Self-financing Condition
Trades are self-financing:
Note that one of the assets can be cash.

7. Implementation Shortfall
Paper portfolio performance over period:
Real portfolio performance over period:

8. Shortfall Calculation
We can rewrite trade term as:
Hence, the shortfall becomes:

9. Cost Components Execution Cost Opportunity Cost Possible set-up:
Adds up shares traded times the difference between the traded price and the reference price (which is the t=0 price).
Opportunity Cost
Adds up overall price moves times the discrepancy between final shares in the real and paper portfolios.
Possible set-up:
Receive new information at t=0 and immediately rebalance paper portfolio. Allow real portfolio until t=T to implement new information. Compare using implementation shortfall methodology.

10. Transactions Costs Forecasting
Commissions and other explicit costs
Easy to model
Typically small for institutional investors.
Spreads
Also easy to observe and model.
Market Impact
Extra cost of trading more than one share.
Liquidity provider assessment of counterparty:
Sheep?
Wolf?
Probabilities vary with urgency of trading.

11. Market Impact Model Inventory Risk
Liquidity provider receives offer to sell Vtrade.
How long before the liquidity provider might receive similar orders (in aggregate) to buy Vtrade?
Risk of holding the position over that time:

12. Market Impact
Liquidity provider demands price concession proportional to risk:
Price impact proportional to square root of shares traded
Market Impact cost proportional to shares traded times cost per share
Proportional to shares traded to power 3/2.

13. Example Trade 10,000 shares: Trade 20,000 shares:
Move price from $50 to $50.25
Price impact = 0.50%
Market impact cost = 10,000 * ($0.25) = $2,500
Trade 20,000 shares:
Price impact
Move price to $50.35
Market impact cost = 20,000 * ($0.35) = $7,000

14. Overall Cost Model
Thomas F. Loeb, “Trading Costs: The Critical Link between Investment Information and
Results.” FAJ, 1983

15. Rule of Thumb Calibration
It costs one day’s volatility to trade one day’s volume.
This sets coefficient c in model.
Calibration with trade data can refine this further.
Improvements on this simple model
Same framework, add structure.
Better forecasts of daily volume and volatility.
Elasticity: as prices move, how does this change the mix of buy and sell orders?

16. Cost-aware Portfolio Construction
We need to adjust our utility function:
Notes:
Costs effectively lower portfolio a.
Utility depends on initial position.
Cost is zero in absence of trading.
th is the tcost amortization horizon

17. Transaction Cost Amortization
The quantities a and w are forecasts over time (i.e. flow variables).
The quantity Cost forecasts costs at the moment of trading.
We need to convert Cost for comparability with a and w .
This number can have a large impact on portfolio construction, yet receives little attention.

18. Dynamics of Portfolio Management
There is something fundamentally wrong with the prior approach.
We are applying one-period technology to an inherently multi-period problem.
The amortization horizon is but one symptom of that.
The true dynamic problem optimizes expected utility over time, accounting for the evolution of alpha and risk over time, and naturally handling cost amortization.

19. Trading and Turnover We define turnover in terms of trading.
Canonical example:
Go from 100% stock A to 100% stock B.
We count that as 100% turnover.

20. Turnover and Utility
Trading costs lower turnover and impact our ability to add value through active management.
We can gain some intuition for the extent of this problem by considering how much value we can add with limited turnover.
We will start with an initial portfolio I, an optimal (ignoring costs) portfolio Q, and consider portfolios P which approach Q as we allow turnover to reach the required amount.

21. Turnover and Utility We define P as:
So P moves linearly from I to Q. We want to work out how utility changes as this happens. And, we will compare the improvement in utility over I, to the maximum available: the improvement of Q over I

22. Maximum Utility Improvement
We achieve it with Q:
And, using first order conditions for Q:
These lead to additional useful relationship:

23. Portfolio P Analysis We calculate: We can simplify this to:
Rule of thumb: we can achieve at least 75% of the added value with 50% of the turnover.

24. Value Added/Turnover Frontier

25. Trading As an Optimization Problem
What is the trader’s challenge?
Move from initial portfolio to final portfolio while keeping costs (market impact) low, risk relative to final portfolio low, and taking advantage of short term opportunities along the way.
Utility Function
What variables can trader control:
Shares traded each day

26. Continuous Single Stock Example
We integrate utility along path
Alpha and risk are straightforward:
New component. We define market impact in proportion to speed of trading
(where )

28. Calculus of Variations
We can solve these general problems. The appendix shows that the optimal solution satisfies:
In this case, the above implies:

29. Single Stock Example

30. Observations Commissions and spreads don’t matter.
We already know how many shares we want to trade. Only issue is how fast.
Risk will motivate faster trading.
Market impact will motivate steady trading.
We need to specify the number of available trading periods.
We usually impose monotonicity constraint.