How to Predict FX Carry Profitability Friday, 13 October, 2017

A new financial research paper related to:

#5 - FX Carry Trade

Authors: Anatolyev, Gospodinov, Jamali, Liu

Title: Foreign Exchange Predictability During the Financial Crisis: Implications for Carry Trade Profitability

Link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3029725

Abstract:

In this paper, we study the effectiveness of carry trade strategies during and after the financial crisis using a flexible approach to modeling currency returns. We decompose the currency returns into multiplicative sign and absolute return components, which exhibit much greater predictability than raw returns. We allow the two components to respond to currency-specific risk factors and use the joint conditional distribution of these components to obtain forecasts of future carry trade returns. Our results suggest that the decomposition model produces higher forecast and directional accuracy than any of the competing models. We show that the forecasting gains translate into economically and statistically significant (risk-adjusted) profitability when trading individual currencies or forming currency portfolios based on the predicted returns from the decomposition model.

Notable quotations from the academic research paper:

"The most signi…ficant departure from the lack of predictability of exchange rates has been documented in the carry trade literature. In a carry trade, an investor borrows in a low-interest currency and invests the borrowed funds in a high-yielding currency. A number of possible explanations have been advanced to account for the positive average returns of carry trade. In a classical asset pricing context, the positive average returns should refl‡ect compensation for bearing a (possibly time-varying) risk premium.

In this paper, we adopt a statistical approach to uncovering and exploiting potential predictability in carry trade returns during and after the recent U.S. …financial crisis. More specifically, we capitalize on the method proposed by Anatolyev and Gospodinov (2010) to decompose currency returns into two multiplicative components (sign and absolute returns) that individually exhibit much greater predictability than raw returns. We then model the joint conditional distribution of these components and use it to produce forecasts of future returns. We allow the two components to respond to currency-specifi…c risk factors such as speculative pressure. This method of incorporating
any implicit nonlinearities in a ‡flexible, indirect fashion is motivated by:
(i) the limited success of linear asset pricing models in explaining carry trade returns
(ii) prior empirical evidence pointing to a statistically and economically signi…ficant element of nonlinear out-of-sample predictability in foreign exchange markets especially at long horizons
(iii) the pro…tability of trading based on the decomposition model of Anatolyev and Gospodinov (2010) for equity returns.

Several interesting results emerge from our analysis.

- First, the decomposition model exhibits substantial directional accuracy in predicting carry trade returns during the recent …financial crisis.
- Second, the out-of-sample forecasting gains of the decomposition model (relative to the naïve historical mean and linear prediction models) translate into economically and statistically highly signi…ficant profi…tability.

We view the uncovered nonlinear predictability as a possible explanation of the limited success of linear asset pricing models in the context of currency markets. Note that the carry trade returns consist of two parts –future currency returns and interest rate differential –and while the pure carry trade strategies exploit only the differential in interest rates, both of which are near the zero lower bound over this period, we employ a model-based carry trade strategy that capitalizes on the predictability of currency returns. We allow the two components (multiplicative sign and absolute return components) to respond to currency-speci…c risk factors and use the joint conditional distribution of these components, modeled as a time-varying copula, to produce forecasts of future returns.

"


Are you looking for more strategies to read about? Check http://quantpedia.com/Screener

Do you want to see performance of trading systems we described? Check http://quantpedia.com/Chart/Performance

Do you want to know more about us? Check http://quantpedia.com/Home/About

Suggestion of a New Currency Factor Model Tuesday, 3 October, 2017

A new research paper related to multiple currency strategies:

Authors: Aloosh, Bekaert

Title: Currency Factors

Link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3022623

Abstract:

We examine the ability of existing and new factor models to explain the comovements of G10-currency changes. Extant currency factors include the carry, volatility, value, and momentum factors. Using a new clustering technique, we find a clear two-block structure in currency comovements with the first block containing mostly the dollar currencies, and the other the European currencies. A factor model incorporating this “clustering” factor and two additional factors, a commodity currency factor and a “world” factor based on trading volumes, fits all bilateral exchange rates well, whatever the currency perspective. In particular, it explains on average about 60% of currency variation and generates a root mean squared error relative to sample correlations of only 0.11. The model also explains a considerable fraction of the variation in emerging market currencies.

Notable quotations from the academic research paper:

"In this paper, we set out to examine various factor models to explain currency comovements and document their fit with the data from a global perspective. That is, we attempt to identify a factor model that works well whatever the currency perspective is. To facilitate a global perspective on currency comovements, we introduce the concept of a “currency basket.” The currency basket simply averages all bilateral currency changes relative to one particular currency. As we show formally, by analyzing 10 currency baskets for the G10 currencies, we span all possible bilateral currency movements. We then contrast the explanatory power of the extant risk factors mentioned previously with the explanatory power of various new factors.

Most importantly, we use a new clustering technique to introduce several new currency factors. When selecting two clusters, a very clear factor structure emerges, with the dollar currencies (Australian, Canadian, New Zealand and US) and the Japanese yen in one block and the European currencies in the other. When using three clusters, a commodity type currency factor also emerges. Combining these statistical factors with a “market” factor, based on currency trading volumes, and a commodity currency factor, we propose several parsimonious factor models and run a horse race versus models incorporating the existing factors.

Among the extant currency factors, the carry and value factors exhibit the highest explanatory power for currency variation. This is not surprising because both factors are relatively highly correlated with the first principal component in bilateral currency rates. However, a new parsimonious factor model incorporating the two-block clustering factor, a commodity factor and the market factor easily beats factor models created from extant risk factors, even models that feature double as many factors. The new factor model explains on average about 60% of the variation in changes in currency basket values. Moreover, the Root Mean Squared Error (RMSE) relative to sample correlations is only about 0.11, which is statistically significantly better than any model based on extant risk factors."
 



Are you looking for more strategies to read about? Check http://quantpedia.com/Screener

Do you want to see performance of trading systems we described? Check http://quantpedia.com/Chart/Performance

Do you want to know more about us? Check http://quantpedia.com/Home/About

Craftsmanship Alpha Thursday, 28 September, 2017

An interesting paper about the artistry in a building of multi-factor portfolios:

Authors: Israel, Jiang, Ross

Title: Craftsmanship Alpha: An Application to Style Investing

Link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3034472

Abstract:

Successful investing requires translating sound investment concepts into actual trading strategies. We study many of the implementation details that portfolio managers need to pay attention to; such choices range from portfolio construction to execution. While these kinds of decisions apply to any type of investment strategy, they are particularly important in the context of style investing. Consider two managers who both intend to capture the value factor in a long/short context: each manager might make a number of decisions, many of which can lead to meaningfully different outcomes. These choices can often explain why one value manager outperforms another. Ultimately, what may seem like inconsequential design decisions can actually matter a lot for style portfolios. In fact, the skillful targeting and capturing of style premia may constitute a form of alpha on its own — one we refer to as “craftsmanship alpha.”

Notable quotations from the academic research paper:

"Style premia are a set of systematic sources of returns that are well researched and have been shown to deliver longrun returns that are uncorrelated with traditional assets. Styles have been most widely studied in U.S. equity markets, but have been shown to work consistently across markets, across geographies, and over time. There are variations in the types of style portfolios, but also — importantly — in how different managers choose to build those portfolios. While practitioners might define styles with similar “labels,” actual portfolios can differ significantly from one another.

Our paper focuses on the craftsmanship required to build effective style portfolios. That is, the kind of decisions that happen after we have already agreed on the type of style portfolio that we want to build.

We start with a brief discussion of the types of style portfolios an investor may choose; we then go into more detail on design decisions related to building style portfolios; and finally, we address other considerations for style investing, such as trading and risk management. We will share our thoughts on a number of enhancements that can be made without deviating from the main thesis . While many of these enhancements reflect our opinions on better ways to build portfolios, the main point is that these choices need to be made consciously. Certain design choices may improve the risk/return characteristics of the overall portfolio, by enhancing returns, reducing risk, or a combination of both. We call the sources of alpha that involve implementation choices “craftsmanship alpha.”

Topics:

1. What Kind of Style Portfolio?
2. How to Build Style Portfolios?
  2.1. Smarter Style Measures
  2.2. Multiple Style Measures
  2.3. Stock Selection and Weighting Schemes
  2.4. Unintended Risks
  2.5. Volatility Targeting
  2.6. Integrating Styles in a Multi-Style Portfolio
  2.7. Strategic or Tactical
3. How To Execute Style Portfolios?
   3.1. Portfolio Implementation
   3.2. Cost-Effective Execution
   3.3. Risk Management
4. Conclusion
"


Are you looking for more strategies to read about? Check http://quantpedia.com/Screener

Do you want to see performance of trading systems we described? Check http://quantpedia.com/Chart/Performance

Do you want to know more about us? Check http://quantpedia.com/Home/About

Global Diversification Works for Multi-Factor Portfolios Sunday, 24 September, 2017

If an investor wants to build multi-factor portfolio then he should look around and build a diversified global portfolio:

Authors: Binstock, Kose, Mazzoleni

Title: Diversification Strikes Again: Evidence from Global Equity Factors

Link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3036423

Abstract:

The benefits of country diversification are well established. This article shows that the same benefits extend to equity factors, such as value, size, momentum, investment, and profitability. Specifically, country factor portfolios reflect both common variation, which we define as the global factor, and local variation. On average, a US investor could enjoy a 30% reduction in portfolio volatility by investing globally. We also document three other properties of equity factors. Like major asset classes, greater market integration is associated with greater factor co-movement, and factor portfolios of different countries tend to be more correlated during bear stock markets. However, unlike asset classes, the correlations of factor portfolios across countries have not been increasing over the last two decades, making global equity factors a particularly desirable addition to a portfolio.

Notable quotations from the academic research paper:

"In this paper, we bring the insights of geographic diversification to cross-sectional equities. We study the returns of long-short portfolios across developed countries based on six factors: market, value, size, momentum, investment, and profitability.

Our international equity factor analysis offers three novel insights.

First, by diversifying an equity strategy across developed markets, investors can significantly reduce the volatility of their factor portfolio. Even for a U.S. investor, who has access to a large domestic market, the volatility reduction across the factors is estimated up to 30%. Indeed, country factor portfolios reflect common variation, which we identify as the global factor, and “local” volatility. A global factor has a simple interpretation as the average world excess returns and tends to explain the individual strategies’ alpha. The local component reflects potentially uncompensated risk, which can be diversified away by simply investing across national markets.

Our second insight shows that factor strategies tend to be more correlated across more integrated countries. For instance, the correlation between the US and the UK stock markets is markedly higher than the correlation between the Japanese and UK markets. We find that these associations also extend to factor portfolios. Accordingly, the momentum strategies in the US and the UK markets are notably more correlated than the momentum strategies in the Japanese and UK markets.

Our last contribution highlights the time-series behavior of factor strategies during bear and bull markets, and across different decades. Previous studies show that return correlations tend to increase in bear markets. Consistent with these works, we document that country factor strategies tend to be more correlated during down-market periods, a phenomenon explained by rising global volatilities. Hence, even for equity factors, diversification fades when most needed. Yet, in contrast to the trends observed for major asset classes, we also document that these correlations have been relatively stable over different decades. This is good news for long-term investors who seek different sources of diversification.
"


Are you looking for more strategies to read about? Check http://quantpedia.com/Screener

Do you want to see performance of trading systems we described? Check http://quantpedia.com/Chart/Performance

Do you want to know more about us? Check http://quantpedia.com/Home/About

Why Machine Learning Funds Fail Friday, 15 September, 2017

An interesting insight into problems associated with an attempts to implement machine learning in trading:

Authors: de Prado

Title: The 7 Reasons Most Machine Learning Funds Fail

Link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3031282

Abstract:

The rate of failure in quantitative finance is high, and particularly so in financial machine learning. The few managers who succeed amass a large amount of assets, and deliver consistently exceptional performance to their investors. However, that is a rare outcome, for reasons that will become apparent in this presentation. Over the past two decades, I have seen many faces come and go, firms started and shut down. In my experience, there are 7 critical mistakes underlying most of those failures.

Notable quotations from the academic research paper:

"
• Over the past 20 years, I have seen many new faces arrive to the financial industry, only to leave shortly after.
• The rate of failure is particularly high in machine learning (ML).
• In my experience, the reasons boil down to 7 common errors:
1. The Sisyphus paradigm
2. Integer differentiation
3. Inefficient sampling
4. Wrong labeling
5. Weighting of non-IID samples
6. Cross-validation leakage
7. Backtest overfitting

Pitfall #1:
The complexities involved in developing a true investment strategy are overwhelming.  Even if the firm provides you with shared services in those areas, you are like a worker at a BMW factory who has been asked to build the entire car alone, by using all the workshops around you. It takes almost as much effort to produce one true investment strategy as to produce a hundred. Every successful quantitative firm I am aware of applies the meta-strategy paradigm. Your firm must set up a research factory where tasks of the assembly line are clearly divided into subtasks, where quality is independently measured and monitored for each subtask, where the role of each quant is to specialize in a particular subtask, to become the best there is at it, while having a holistic view of the entire process.

Pitfall #2:
In order to perform inferential analyses, researchers need to work with invariant processes, such as returns on prices (or changes in log-prices), changes in yield, changes in volatility. These operations make the series stationary, at the expense of removing all memory from the original series. Memory is the basis for the model’s predictive power. The dilemma is returns are stationary however memory-less; and prices have memory however they are non-stationary.

Pitfall #3:
Information does not arrive to the market at a constant entropy rate. Sampling data in chronological intervals means that the informational content of the individual observations is far from constant. A better approach is to sample observations as a subordinated process of the amount of information exchanged: Trade bars. Volume bars. Dollar bars. Volatility or runs bars. Order imbalance bars. Entropy bars.

Pitfall #4:
Virtually all ML papers in finance label observations using the fixed-time horizon method. There are several reasons to avoid such labeling approach: Time bars do not exhibit good statistical properties and the same threshold 𝜏 is applied regardless of the observed volatility. There are a couple of better alternatives, but even these improvements miss a key flaw of the fixed-time horizon method: the path followed by prices.

Pitfall #5:
Most non-financial ML researchers can assume that observations are drawn from IID processes. For example, you can obtain blood samples from a large number of patients, and measure their cholesterol. Of course, various underlying common factors will shift the mean and standard deviation of the cholesterol distribution, but the samples are still independent: There is one observation per subject. Suppose you take those blood samples, and someone in your laboratory spills blood from each tube to the following 9 tubes to their right. Now you need to determine the features predictive of high cholesterol (diet, exercise, age, etc.), without knowing for sure the cholesterol level of each patient. That is the equivalent challenge that we face in financial ML.
–Labels are decided by outcomes.
–Outcomes are decided over multiple observations.
–Because labels overlap in time, we cannot be certain about what observed features caused an effect.

Pitfall #6:
One reason k-fold CV fails in finance is because observations cannot be assumed to be drawn from an IID process. Leakage takes place when the training set contains information that also appears in the testing set. In the presence of irrelevant features, leakage leads to false discoveries. One way to reduce leakage is to purge from the training set all observations whose labels overlapped in time with those labels included in the testing set. I call this process purging.

Pitfall #7:
Backtest overfitting due to data dredging. Solution - use The Deflated Sharpe Ratio - it computes the probability that the Sharpe Ratio (SR) is statistically significant, after controlling for the inflationary effect of multiple trials, data dredging, non-normal returns and shorter sample lengths.
"


Are you looking for more strategies to read about? Check http://quantpedia.com/Screener

Do you want to see performance of trading systems we described? Check http://quantpedia.com/Chart/Performance

Do you want to know more about us? Check http://quantpedia.com/Home/About