Quality Money Management : Process Engineering and Best Practices for Systematic Trading and Investment

144 CHAPTER ◆ 1 4 Gather Historical Data

14.2.5. Economic Data

Economic data, such as CPI and GDP, are key indicators and can alert the product team

as to when a trading/investment style is or is no longer working. For example, some strat-

egies, say small cap value, tends to work quite well coming out of a recession, because

small companies are hard to value. In a recession, properly valuing small companies can

generate large returns. Large cap momentum strategies tend to work very poorly, because

you tend to buy companies that have outperformed the index the most, which tends to

be lagging stocks. The lagging stocks tend to be the next group to decrease in price.

Economic data is primarily for trading/investment systems that:

● Make country selections or bets on currencies.

● Select sectors or capitalizations.

● Calculate a value/growth mixture.

● Build Bayesian-style indicators to shift weights to those that perform better under

particular economic conditions.

Broad economic factors can be used to shut down trading/investment systems when the

economics no longer justify the strategy. When the economic cycle shifts and depending

upon the nature of the trading/investment system, the product team may stop trading, or

close the fund, as opposed to continuing to trade and losing money in a bad environment.

SPC will raise a flag when something has fundamentally shifted. (We use economic

data for design of experiments, to look for assignable causes, except for currency trading

systems where the main signals are economic data.) By using SPC to analyze the outputs

of the trading algorithms, the product team can detect a shift in the underlying stochastic

process. Once detected in a backtest, the team can then perform root cause analysis to

determine what caused the shift. The first step of root cause analysis should be an analy-

sis of economic cycle data around the time the algorithm stopped working (e.g., did our

growth algorithm stop working three months before a recession?).

14.3. STEP 1, LOOP 2: Purchase Data

What a trading/investment firm buys from a data vendor is not just raw data; it buys engi-

neering and capability. Once the data requirements are firmly established over the course

of the previous loop. To get correct data, the team should produce a request for quotation.

A quotation from a data vendor should answer several questions in addition to the cost.

1. How do we get the data? Streaming, real-time data feed via API or Excel DDE
   Link? FTP open or closed?


2. How much responsibility does the vendor take for the data?
   (a) Who does the splits?
   (b) Who synchronizes the data?
   (c) Who stores the historical data?
   (d) Who normalizes the data?


3. Which frequency do we use? It should be in the frequency we expect to trade in.
   (a) Trade-by-trade tick data?
   (b) Time intervals or bars?