costs for UK property. Table 1 shows the ranges for the transaction cost components according to The Global Property Guide 2007. Looking for boundary values, we take the maximum of the ranges. The maximum value of the buyer’s transaction cost range k_{1b }is 5.15% while the maximum value of the seller’s range k_{1s }is 4.11%.

[Insert Table 1 about here]

To assign a value to the frictions k_{2 }and k_{3 }we fit both the upper and lower bound to observed prices. In particular, we solve for the value of the frictions such that the bounds are as narrow as possible but all spread observations p^{∗ }just remain within the arbitrage free price bounds, i.e.

m i n k 3 2 , k [ p ( k 2 )

p(k_{2}, k_{3})]

(18)

subject to

p(k_{2}, k_{3}) ≤ p^{∗ }≤ p(k_{2}).

(19)

This process leads to a market implied cost of 4.46% for transaction time k_{2 }and one of 11.83% for the short sale constraint k_{3}. Fig. 5 plots the historical trajectory of the Halifax HPI and the arbitrage free price bounds for its forward prices, using the obtained values for the market frictions.

[Insert Fig. 5 about here]

# Veri cation of friction costs

The next step is to verify the obtained values for the implied friction costs k_{2 }and k_{3}.

The cost of transaction time k_{2 }is a measure of marketability for illiquid assets. Longstaff (1995) derives an analytical upper bound on the value of marketability using option pricing theory. Dyl and Jiang (2008) use the Longstaff model to value illiquid common stock. The model requires only two inputs: the volatility of the considered asset and the length of time the asset is illiquid or the time it takes to sell the illiquid asset. The model’s fundamental insight is that marketability is properly construed as the option to sell an asset at the time of one’s own choosing. In particular, Longstaff uses a lookback option that pays the difference between the maximum asset value during the marketability

11