Tradable economics is a technology for building systematic trading strategies based on economic data. Economic data are statistics that – unlike market prices – directly inform on economic activity. Tradable economics is not a zero-sum game. Trading profits are ultimately paid out of the economic gains from a faster and smoother alignment of market prices with economic conditions. Hence, technological advances in the field increase the value generation or “alpha” of the asset management industry overall. This suggests that the technology is highly scalable. One critical step is to make economic data applicable to systematic trading or trading support tools, which requires considerable investment in data wrangling, transformation, econometric estimation, documentation, and economic research.

The below post is based on conversations with a range of experts on the subject and on direct experience with this technology at Macrosynergy Partners.

The post ties in with the SRSV summary on quantitative methods for macro information efficiency.

What is tradable economics?

Tradable economics here refers to a technology that deploys economic data in systematic trading strategies. Economic data here mean statistics that inform directly on economic activity. Financial market prices may be informative on economic activity, but only indirectly and in dependence upon the functioning and information efficiency of the market. Economic data comprise official economic statistics (as released by national statistics offices, central banks, international organizations or industry associations), survey data (typically published by research companies or institutes), corporate balance sheets, alternative data (such as satellite images, internet activity, financial transactions, mobile communication) and data produced from natural language processing (view post here).

Published economic data are informative for financial market trading but typically cannot be easily and directly plugged into systematic trading strategies. Unlike financial market data, which are intensively used for algorithmic and systematic trading, economic data come with a number of inconvenient features such as low frequency of updating, lack of point-in-time recording and backward revisions. Therefore, tradable economics typically operates in two distinct stages:

  • First, economics statistics and other quantifiable information must be brought into a form that is suitable for systematic research. One can call this form tradable economic data.
  • Second, these tradable economic data must be integrated into a quantamental system that supports cost-efficient research and development of systematic trading strategies or trading support tools. On this point view post here.

Typically, economic developments, such as business cycle phases, inflation trends or external imbalances prevail over months or even years. Hence, by the standards of financial market trading strategies most economic developments are low-frequency events. Only a limited number of economic states or trends have been observed for each country over the past few decades of available data. Thus, rather than just relying on the power of large data sets, tradable economics must lean more heavily on theoretical priors. It is important to integrate plausible and compelling propositions of economic theory with the available economic data.

Why is tradable economics a big deal?

Tradable economics offers huge potential. While it is still in its infancy today, it is highly scalable globally across the asset management industry with a positive impact on aggregate returns.

Today many popular trading strategies are zero sum games that require traders or trading systems to outsmart others. This preference is not surprising: experienced traders know each other and the financial industry better than they know the economy at large. All strategies that are based on market positioning, market flows or anticipated algorithmic rebalancing belong into that category. Even some types of trend following are partially based on exploiting and exacerbating market distress (“pushing prices through stops”). In these cases, investor return is created by one financial institution at the expense of another. The aggregate alpha that is generated by the asset management industry through such trading strategies does not principally increase with technological advances.

For example, a trading style based on under- or over-positioning of the market is a form of legitimate front-running. It will turn a profit if one manager’s judgment on positioning is better than that of others. The profit does, however, come at the expense of higher position entry or exit costs for other managers.

This is different for tradable economics. The principal value generation in tradable economics arises from efficiency gains that come with a better alignment of market prices and economic conditions. Economic alignment creates two benefits for the overall financial system. The first is more efficient allocation of resources; prices become more meaningful and guide flows of goods, services, and investments based on better information. Put differently, economic alignment of market prices reduces the risk of severe and enduring misallocation and hence of major economic and financial crises. The second benefit is faster and more reliable feedback to actions taking by policymakers or managements of public companies; harmful decisions are more easily discovered and more quickly reversed.

For example, macro traders that figure out early that economic growth is declining in a country would – in a simple case – receive rates and short the currency of that country until a plausible adjustment to the new economic position has been reached. Less well-informed managers may or may not benefit from this adjustment but they do not necessarily incur a systematic loss because of the activity of the informed manager as the adjustment would ultimately have taken place anyway. The positive PnL of a tradable economics strategy is effectively paid from the economic gain of the majority of economic agents that benefit from the market price adjustment and can be regarded as a payment for expediting that adjustment.

Since tradable economics is not a zero sum game, it is a scalable technology for the investment management industry overall. Technological advances in tradable economics can increase the performance of the asset management industry as a whole and – in particular – the performance of active manager vis-à-vis passive investment vehicles.

How do economic data conventionally feed into the investment process?

Few people would seriously dispute the importance of economic data for market prices. Economic data already play a prominent role in financial market trading, particularly in the macro space. Two common examples are data release trading and the use of economic forecasts in asset valuation models.

  • Data release trading means that traders take positions before and after an economic report is released based initially on their expectations and, subsequently, their interpretation of the published number (and the markets’ attitude towards it). The informed trader will have a better forecast, a better sense of market positioning and a better understanding of the actual meaning of the number. A particular attraction of this type of trade is the predictability of market attention: a significant part of the market will consider and respond to the economic information in a specific time interval. This keeps the required time span for the risk exposure short.
  • Incorporating economic predictions into valuation models is also quite common. For example, expectations for nominal GDP growth support corporate earnings estimates and term premia estimates in bond markets. Usuallly the economic forecasts are taken from economists’ published work and are updated infrequently and sporadically.

Financial markets have been less successful in using macro data continuously and systematically for trading. Indeed, the vast majority of macro systematic strategies do not use actual economic data but rather market prices that have some economic interpretation or derivatives thereof. For example, commodity price-bases terms-of-trade indices are used for foreign exchange trading, corporate CDS spreads are used for equity trading, and real exchange rates are used for fixed income positioning These systematic strategies rather arbitrage information across different markets rather than bringing economic information into financial markets.

How to generate tradable economic data for systematic strategies?

Conventional economic statistics are not suitable as direct input for trading models. The most obvious obstacles are:

  • Time series are not recorded in ‘point-in-time’ form. Point-in-time means that economic information is assigned to its release date in the form it was released. Instead, the standard format of economic time series aligns latest revised information with reference period of the underlying economic activity. Reconstructing the point in time format can carry considerable costs.
  • Most economic data series are impaired by breaks in the series, changes in definition and scope, and outlier distortions. These problems are pervasive across countries and types of data series, but particularly pronounced for specific series, such as credit and government finance data where legal changes and one-off balance sheet transactions can have a pronounced impact.
  • Many economic data are hard to interpret with respect to what they actually measure when they measured it and what adjustments have been applied. In some emerging countries, data even have deliberately been tampered with in the past. Some alternative data and survey providers can be somewhat disingenuous in the presentation of their data. Without a deep understanding of conventions, most data reports produce more noise than signal for economic trends and some can be outright misleading.

This means that unlike in the case of market price data, economic data need a lot of detailed background knowledge and experience that lies outside the core business of financial institutions. Without such knowledge, even the most elaborate statistical transformation will not be able to transform the series appropriately. In order to allow researchers to develop theoretically sound and robust trading strategies tradable economic should provide the following features:

  • Intuitive interpretation: the conceived object of a tradable economic data series should have clear meaning for investors. This could be something like broad economic or GDP growth, aggregate demand growth, popular inflation expectations and so forth. A conceptually clear measure must often be calculated based on a set of published data series and an appropriate form of aggregation. The exact definition and documentation is very important, so that meaning is not confused “downstream” in the development stage of trading strategies.
  • High quality: the tradable economic data should be adjusted for all effects that distort the numbers relative to the concept that they are meant to represent. This typically includes adjustments for seasonal effects, working days, holiday patterns, tax effects and many other aberrations. There are statistical methods (and packages) that help the process but without expert oversight and curation they are not fully effective.
  • Econometric estimation: many conceptual economic time series are not directly observable but need to be estimated. This includes for example output gaps, inflation expectations or cyclically-adjusted fiscal deficits. Thanks to the advances in statistical programming, particularly through the R project and Python libraries, econometric production of meaningful economic data has become a lot cheaper and easier in past years (view post here, https://macrosynergy.com/the-power-of-r-for-trading-part-2/).
  • Point-in-time recording: for any point in time a recorded statistic should only contain information that was publicly available at that moment, whether this is actual data or estimated parameters. True time series of available economic information condense many different vintage series to record the evolution of information over time. Technically, the point-in-time format updates a complete history across time. It requires the full set of information available at each release dates. There are some off-the-shelf point-in-time economic databases (ALFRED archival economic data of the St. Louis Fed, OECD revision database), but they are not usually sufficient for professional asset management.
  • High frequency: tradable economics should have a daily frequency at least for the convenience of strategy developers. Higher frequency would be desirable if release-related strategies are considered. For single-release series, this just results in infrequent updating at times of release and revision. But tradable economic data based on multiple series may naturally update almost daily, albeit with different variability depending on what data are released. An example would be the daily U.S. real-time recession risk tracker of Jesse Edgerton at J.P. Morgan.

A strong case can be made that tradable economic data (unlike tradable economics strategies) is a scale business, meaning that producing such data in larger quantities would result in disproportionately higher trading profits per dollar invested. This reflects fixed costs in producing such data alongside with disproportionately growing opportunities with the breadth of available tradable economic data series:

  • A key fix cost is the wrangling and transformation of economic data series. This requires specialized packages (for example for R or Python), specialized databases (of less well-publicized information such as release dates, vintage data series, school holiday patterns, methodology changes) and export consultations (as almost every type of economic data in each country has idiosyncratic issues). Once these structures are in place they can be applied for a broad and global range of economic statistics.
  • Tradeable economics strategies typically require some combination of tradable economic data series and some of these series may have to be customized. This means that the strategy developer needs a broad range of tradable data series to research and use for development. Indeed, if the strategy uses supervised learning it necessarily requires a range of plausible metrics for one and the same conceptual trading factor so that the algorithm can choose or weigh the alternatives.
Share

Related articles