While economic information undeniably influences financial markets, incorporating it into systematic trading strategies has faced limitations. Skepticism towards economic theory, data quality issues, and a lack of readily usable formats all contributed to this. However, advancements in statistical learning methods and the emergence of industry-wide quantamental indicators are making macroeconomic data a more powerful tool for traders.
Get your free trading tech book PDF
Challenges of Using Macro Data for Trading
The potential benefits of incorporating macroeconomic information into trading strategies are clear. Economic theory suggests market prices reflect a broader macroeconomic equilibrium, influenced by economic states and shocks. Additionally, discretionary trading based on these fundamentals has a proven track record. However, two major obstacles have hindered the systematic use of macroeconomic data:
Get your free trading tech book PDF
Indirect and Complex Relationships: The connections between economic data and market prices are often indirect and intricate. Building trading signals requires sound judgment based on macroeconomic theory and data analysis. However, portfolio managers and trading system engineers may not possess strong expertise in macroeconomics, and economists often lack consensus views.
Data Quality Issues: Utilizing macroeconomic data for trading presents challenges due to several data quality issues:
Sparse History: Many economic data series, particularly in emerging economies, have limited historical coverage. This restricts the ability to capture the full range of business cycles and financial crises. Combining data from multiple regions and stitching together different series might be necessary.
Revisions: Economic databases store data in their latest revised form. Initial releases of indicators like GDP or business surveys may differ significantly. These revisions account for updated information, methodological changes, and adjustments for seasonal and calendar effects. Essentially, the information recorded for the past may not reflect what was available at the time.
Dual Time Stamps: Unlike market data, economic records have separate observation periods (when the event occurred) and release dates (when the data becomes public). Databases only associate values with observation periods.
Distortions: Most economic data are subject to temporary distortions relative to what they aim to measure. For example, inflation data can be affected by one-off tax changes or price hikes. Production and balance sheet data can reflect disruptions like strikes or unusual weather. Additionally, there can be sudden breaks in time series due to methodological changes. Occasionally, data may be deliberately manipulated for political reasons.
Calendar Effects: Many economic data series are significantly influenced by seasonal patterns, working day variations, and school holiday schedules. While some series are calendar-adjusted, others are not, and these adjustments often lack consistency across countries.
Multicollinearity: The variations of many economic data series are correlated due to common factors like business cycles and financial crises. This often means that multiple data points tell a similar story. Extracting latent factors that represent these common trends requires domain knowledge, statistical methods, or a combination of both.
Data wrangling, the process of transforming raw data into clean, usable formats, is particularly complex for macroeconomic trading indicators. It involves:
Adapting Indicators for Trading: This requires transforming activity records into information about market states. Common techniques include stitching together different series across time, combining updates and revisions into “vintage matrices” to represent a single point-in-time series, and assigning publication timestamps to periodic updates and revisions of time series.
Filtering and Adjustments: Economic information typically involves filters and adjustments. The parameters for these filters need to be estimated sequentially without introducing bias from looking ahead. Standard procedures include seasonal, working day, and calendar adjustments, special holiday pattern adjustments, outlier adjustments, and flexible filtering of volatile series. Seasonal adjustment is often handled by specialized software, although modules are available in R and Python for programmatic access.
Accounting for Market Perceptions: Markets often interpret information through the lens of economists. To track these analyses over time, it’s crucial to account for changing models, variables, and parameters. Machine learning methods can replicate a plausible evolution of economic analysis. Conventional econometric models are not suitable for backtesting because they are built with hindsight and don’t aim to replicate actual past trends. Machine learning can simulate changing models, hyperparameters (tuning parameters), and model coefficients. One practical approach is “two-stage supervised learning.” The first stage involves scouting features, while the second evaluates candidate models and selects the one that performs best at any given point in time. Another example is simulating the results of “nowcasters” (methods for estimating current information states) over time. This method estimates past information states through a three-step approach: variable pre-selection, orthogonalized factor formation, and regression-based prediction.
News Analysis: News and commentary significantly influence asset prices, arguably more so than traditional price and economic data. However, the vast amount of verbal information is overwhelming for human analysis. Natural language processing (NLP) technology is becoming increasingly important for quantitative evaluation of textual
This article was summarized from
macrosynergy.com/research/macroeconomic-data-and-systematic-trading-strategies/
Comments