March 02, 2011
The portfolio model input
The model’s inputs are
proprietary individual country indicators whichmeasure the underlying economic strength (contraction or expansion)of 10 currencies: NZD, AUD, CAD, JPY, EUR, GBP, USD, CHF, SEK, andNOK. The country indicators, derived from public macroeconomic data,are designed to reflect the macroeconomic strength of each economy.The allocation signals are generated by changes in spreads betweenthe fundamental country indicators. More capital is allocated tocurrencies with relatively strong economic activity (and positive rateoutlook), funded by short positions on currencies with weak economicactivity (weak rate outlook). For example, if the Eurozone fundamentalcountry index suddenly drops (increases) relative to the USfundamental index, the model, all else being equal, would reduce(increase) exposure to EURUSD. Additionally, positions are scaled upor down according to the volatility of the currency crosses in questionso the expected risk-adjusted return for positions in EURCHF is thesame as for positions in the normally more volatile EURCAD.Allocations are presented as net exposures against EUR, USD, or GBPto reduce both the number of possible combinations and most illiquidcrosses.Returns are based on Bloomberg monthly carry-adjusted currencydata. The model therefore does not include costs related to minimumtrading size, slippage, rollover, spreads, and taxes.
The model will be published on www.tradingfloor.com by Saxo Bank onthe first banking day of the calendar month. While Saxo Bank
publishes the model’s suggested allocation,
the bank is not responsiblefor the monthly reweighting of the portfolio.The net exposure of a portfolio does not necessarily equal the nominalportfolio amount. As an example, in a EUR-denominated account thesum of all EUR positions following the model can deviate from theamount allocated to follow the model. Assuming the holder of a EUR 1million account might choose to allocate EUR 1 million to follow themodel, but the sum of EUR exposure will not equal EUR 1 million. Thereason is that one needs to look at the net exposures. If the model islong 100,000 EURUSD and short 100,000 EURJPY, the net exposure inEUR on these two positions is actually zero. The sum of total positionsizes in EUR might therefore deviate from EUR 1 million, since themodel is only looking at net exposures of the currencies in question.The reason is that the model follows 10 currencies, but the netexposures are established via only nine crosses. The sum of all theseexposures is then either net long or short, depending on the m
prediction on EUR itself.