Most top-down and global macro managers follow all of the macro data announced each day. They will compare data across time and countries to understand relative economic performance, but when they go back to basics to understand how data are constructed, they will usually get a very uneasy feeling. How much noise is in this data?

The construction of economic data is an ugly process. If you don’t believe me, read any report on how data is collected. See Diane Coyle’s GDP as a short read, or a recent podcast. This problem becomes especially acute when economists are trying to measure potential GDP and GDP gaps. For that matter, it can be a problem for any cross-country comparisons.

There are delays with data. There are errors with the collection of data. There are data revisions. There are simplifications and seasonal adjustments of data. With surveys, there are limited sample sizes. Yet, traders will often react to any small surprise in the numbers. Traders will react to the actual number. They will react to the surprise as measured by difference between actual and expected. They will react to revisions. It is a tough game to play because data from one series today may be contradicted with data tomorrow from a different series.

There is an alternative to trying to use noisy fundamental signals. Use price data as the primal or core driver for decision-making. Prices provide core signals of market views and expectations and are updated in real time.

The use of price data versus fundamental is a trade-off with signal to noise.  In the case of macro data, there is a signal surrounded by the construction noise. This macro data noise can be smoothed through times series techniques or multi-signal aggregation. In the case of prices, there is a signal surrounded by the noise of non-information driven trading. The price noise can be smoothed through time series techniques.














However, there is a key difference. Any signal to noise analysis with macro data is a two-step process. There is the smoothing of signal to noise with the macro data and then there is the signal to noise link between the macro data and prices. In the case of price trend data, there is only a one-step process of linking past price trends with future prices. There is a deeper hurdle to be overcome with macro data. It may be worth it, but it should always be compared against the simpler process of extracting signals from prices.

Neither may be perfect but prices and their relationship across markets can be viewed as primal.