Understanding macro information (in-) efficiency

The importance of macro information efficiency

Macro information efficiency here means that financial markets produce and use economic research such that the resulting incremental investment returns and social benefits exceed costs. Macroeconomic information here means all public information that are significant for country, asset-class, and sector wide price changes. This information includes economic performance, government and corporate balance sheets, financial market structure and prices, social and political developments, and even environmental and weather trends.

Financial market prices are a key part of the global macroeconomic equilibrium:

• Interest rates affect agents’ savings, consumption and investment decisions.
• Credit spreads influence the conditions of lending to and borrowing of households, firms and sovereigns.
• Exchange rates affect the competitiveness and net asset positions of countries.
• Equity prices determine the attractiveness of taking entrepreneurial risk.
• And commodity prices shape the terms-of-trade of economies and companies.

In the context of macro trading the main social value of macroeconomic information efficiency is the alignment of prices with economic conditions. This means that macro traders can foster the efficient allocation of resources across the world economy. At the same time portfolio managers that invest in macro research can create investor value by detecting fundamental economic and price trends early, before the more sluggish part of the market has fully adjusted to a changing environment.

Why markets are not (macro) information efficient

Information technology is no guarantee for information efficiency. Evidently, the quantity of economic and financial data has surged. Also, transmission speed, scope, and detail of non-quantitative information have expanded. Databases, analytical software, and programming languages for dealing with large quantities of information have advanced alongside. Yet, after all this progress, there is reason and evidence to suggest that markets have not come even close to exhausting their potential for information efficiency.

• The seminal article “On the Impossibility of Informationally Efficient Markets” published by Grossman and Stiglitz in 1980 explains that since price-relevant information is costly it will only be procured to the extent that inefficient markets allow translating it into sufficient returns. As the article points out: ” The only way informed traders can earn a return on their activity of information gathering, is if they can use their information to take positions in the market which are ‘better’ than the positions of uninformed traders…Hence the assumptions that all markets, including that for information, are always in equilibrium and always perfectly arbitraged are inconsistent when arbitrage is costly” (view journal article here).
• Theoretical research suggests that even (fundamental) value traders only invest in research and information if (i) information cost is not too high, (ii) overall markets are poorly informed, and (iii) market makers do not suspect that value traders are well informed (view post here). Point (ii) is important, because the information content of prices with respect to known fundamentals tends to grow faster than that of private research and, hence, often discourages such research (view post here). Point (iii) matters because a value trader with a reputation of being well informed is easily ‘front run’ when giving orders to market makers.  In summary, investment in information is a complex trade-off, with no guarantee that markets as a whole set asset prices close to their fundamental value.
• Experimental research has also produced robust evidence for mispricing of assets relative to their fundamental values. Academic studies support a wide range of causes for such mispricing, including asset supply, peer performance pressure, overconfidence in private information, speculative overpricing, risk aversion, confusion about macroeconomic signals and – more generally – inexperience and cognitive limitations of market participants (view post here).
• In practice, macroeconomic and financial research is tedious and expensive and often beyond the scope of portfolio management teams even at large institutions. This reflects two basic difficulties.
  • First, economic data are notoriously hard to interpret and require considerable adjustments. Forecasters struggle with separating signal from noise and cannot update their predictions continuously. Forecasts are also not easily comparable across countries due to different convention and biases. All this gives rise to “rational information inattentiveness of markets” (view post here). There is empirical evidence of expectation stickiness that goes some way in explaining price momentum after important relevant news, such as corporate earnings releases (view post here). Even the most popular and highest-quality economic data, such as U.S. labor market reports, need in-depth research to be understood (view post here).
  • Second, understanding and researching the relation between economic information and asset prices requires profound experience and some econometric skill. Data science has come a long way in providing powerful tools for analysis and model construction (see final section of this page). However, in the data-constrained macro space the success of statistical models still hinges on good judgment and real in-depth understanding of methods, models and data, which remains in short supply. Most, institutional investors prefer simple relations, often condensed in the three main categories of risk premia strategies, i.e. carry, momentum, and relative value (view post here).
• Most importantly, the diffusion of information through markets depends on transactions. It is not enough to research and publish. Financial markets research translates into price information only if it is acted upon. In practice, the transformation of research and information into positions is often obstructed by institutional rules and regulations. Also, for some institutions market access is limited and trading costs can be prohibitively high. For example, in OTC markets bid-offer spreads vary across clients (view post here). Moreover,  many investment managers simply do not fully trust their researchers. Finally, there is evidence that financial decision-making under uncertainty is far from rational and subject to a range of behavioral biases, such as illusion of control, anchoring bias, sunk-cost bias and gambler’s fallacy (view post here). Altogether, “obstruction to diffusion” is common and often means that relevant information and research translate into gradual price trends rather than instantaneous price adjustments (view post here).

The above points go some way in explaining why asset reallocation often just follows past performance (view post here) and why portfolio managers often find it more practical to generate “fake alpha” through receiving risk premia on exposure to non-directional conventional factors and strategies rather than to generate true investor value (view post here). Information inefficiency also explains why simple trend following has been a profitable trading strategy, even in the best researched and most liquid markets (view post here). Moreover, there is ample evidence of herding and sequential dissemination of information in markets with great macroeconomic importance, including currencies (view post here, and here). Above all, these phenomena testify to the sluggishness of market responses to broad shifts in fundamental conditions.

A useful framework

Efficiency based on macro factors

Investment managers can contribute to and benefit from information efficiency. A simple and practical approach is [i] to create indicators with meaningful macroeconomic and market information and [ii] to condense them into meaningful conceptual macro factors that can guide specific investment strategies. The usage of macro factors has great benefits

• Macro factors make quantitative information manageable. In international macro trading there are way too many economic data series to keep track of, even for the most diligent investment manager. Selected pre-filtered macro factors effectively outsource part of the production of information to research.
• Macro factors support consistency. Often enough investors are myopic: they overrate that latest “fashionable” factors that happened to coincide with recent market moves, regardless of causality and long-term relations. This constitutes a form of “overfitting” of information (view post here). This “overfitting” can lead to the misinterpretation of fundamental information, a phenomenon that has been labelled “scapegoat theory” view post here.
• Macro factors are applicable across strategies. Detecting changes in economic growth, for example, matters for both equity and FX strategies. If returns across asset class strategies have little or now correlation (view post here on equity and FX), macro factors can become important building blocks for diversified multi-strategy portfolios.

Using more than one macro factor in a single strategy is a major challenge. In dealing with different ideas or signals, investment managers veer towards “rules of thumb” of dubious origins, while academic researchers veer towards complex mathematical models. The former are evidently inefficient, while the latter are usually intractable. Hence, for practical purpose and in order to avoid the worst excesses of double-counting, mis-interpreting and forgetting information it is helpful to structure macro factors into three groups.

• Valuation gaps are defined as differences between the market price of an asset or derivative and its estimated value. There are two basic methods of tracking valuation gaps. The first is to estimate an asset’s fundamental value directly, maybe based on discounted cash flows, and then compare it with the quoted price. The second method works indirectly, by efficiently “nowcasting” the trend in key valuation-relevant factors and then estimating the gap between this trend and market perceptions.
• Implicit subsidies are defined as premia paid or discounts offered to financial investors by market participants other than those that maximize risk-adjusted returns. Typically, subsidies are paid by central banks, governments, highly-regulated institutions or non-financial institutions.
• Setback risk is defined as the probability of a mark-to-market drawdown on a position for reasons that are unrelated to fundamental value. Related macro factors typically measure the positioning or “crowdedness” in a trade as well as the probability that investors will exit the trade in the near term.

Importantly, these three types of indicators are complementary, not competing. Indeed, it is not wise to use them in isolation. Thus, a price-value gap often arises as consequence of implicit subsidies: the subsidized asset becomes overpriced for a reason. Also, typically setback risks arise alongside subsidies, causing sudden large losses to subsidy receivers.

This site also presents a fourth principle of value generation, based on the research of price distortions. This principle is explained in a separate summary page (view here) because the exploitation of price distortions is not primarily an issue of macro information efficiency.  Value generation based on apparent price distortions typically depends on the investor’s basic understanding of market flows, easy access to markets and funding, and freedom to operate without the constraints and conventions that trap many institutional investors.

Support from data science

The advances in the science of “statistical learning” have enhanced researchers’ ability to work with large data sets, through use of a wide variety of methods ranging from simple regression to complex machine learning (view post here).

• For example, structural vector autoregression is a most practical model class that helps identifying specific market and macro shocks, with a relative small set of assumptions on market and economic structure (view post here).
• Machine learning partly automates the construction of forecast models through the study of data patterns, the selection of best functional form for a given level of complexity and the selection of best level of complexity for out of sample forecasting (view post here). This includes statistical methods for condensing the information value of the vast multitude of macroeconomic time series, making them more manageable for forecasting, such as Dynamic Factor Models of Sufficient Dimension Reduction (view post here). Similarly, there are regression-based variable selection tools such as “LASSO” or “Elastic Net” (view post here). Beyond mere convenience, these also allow backtests that are based on methods rather than on specific factors whose choice may be biased by historical experience. This gives us a better idea how our way of thinking about macro strategies would have translated into trading profits in the past.