Return Decomposition of Hybrid Strategies in the Brazilian Market: Evidence of Genuine Alpha and Factor Exposure

arXiv:2611.06919 · q-fin.ST, q-fin.PM · Submitted 2026-05-09 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Quantitative Finance. The day in quantitative finance: models of markets, risk and prices.

Ruth: We're looking at the paper: "Return Decomposition of Hybrid Strategies in the Brazilian Market".

Warren: Here's the gist of the paper. This paper decomposes the return sources of a hybrid portfolio combining a fundamentalist screen with a moving-average crossover system using simple moving averages to evaluate genuine alpha and factor exposure in the Brazilian market The central conclusion is unequivocal: the portfolio generated a Jensen alpha of 1.11% per month (p = 0.005) in the Fama-French three-factor model Portfolio Construction and Methodology The study employs a two-layer methodology to investigate return decomposition In the first layer, a portfolio was built through a rigorous fundamentalist screen using ex-ante available data The selection parameters included (i) P/E ratio ≤ 15, (ii) ROE > 15%, (iii) Net Margin > 15%, (iv) Current Ratio ≥ 1, and (v) P/BV between 0.5 and 1.5 This screen was applied once at the beginning of the analysis period, generating a fixed set of 13 assets without periodic rebalancing The operationalization involved a trend-tracking algorithm based on a simple moving average crossover system with SMA of 9 and 46 periods Factor Exposure Analysis The econometric models were used to decompose return sources and test for alpha generation The CAPM model showed a statistically significant alpha of 1.18% per month (p = 0.016) The Fama-French three-factor model also revealed a positive and statistically significant Jensen's alpha of 1.11% per month (p = 0.005) The SMB factor was highly significant with a coefficient of βs = 0.3808 (p = 0.000) Conversely, the HML factor was not statistically significant (βh = 0.1266; p = 0.544) The four-factor model with momentum confirmed the robustness of the finding, showing an alpha of 0.98% per month (p = 0.009) Performance and Risk Assessment Descriptive metrics indicated that the hybrid portfolio posted an average net profit of BRL 10.2 million over 25 years, exceeding the benchmark average The Sharpe ratio for the hybrid portfolio was 0.443 versus 0.039 for the Ibovespa, although this difference was not statistically significant under the Jobson-Korkie test (p = 0.154) The average drawdown of the hybrid portfolio was 25.49%, which is substantially higher than the benchmark's 10.

Title and authors: Ruth: So we're looking at this paper now, "Return Decomposition of Hybrid Strategies in the Brazilian Market: Evidence of Genuine Alpha and Factor Exposure." The title tells you right away that they aren't just looking at one strategy; they are trying to separate what’s genuine skill from what’s just market noise or known factors.

Warren: Exactly. It combines a fundamentalist screen with a moving average crossover system, so the core question is whether that hybrid approach actually finds new returns or if it’s just riding on some well-known risk premia in Brazil.

Iris: The authors are Francisco Robson do Nascimento de Freitas, Felipe Macedo Zumba, Lucio Marassi, and Jian Vingthor Souza Luna de Azevedo from UFRN. They brought together different viewpoints here, which is important because they’re using a mixed methodology to get a clearer picture of the return sources.

Ruth: Right. And what makes this work interesting for us is that they are specifically testing this combination in the Brazilian stock exchange, B3, over a long period from two thousand to two thousand twenty-five <ref:2611.06919#pg1,the Brazilian stock exchange, B3>. It’s not just theoretical stuff; it’s applied empirical evidence from a real market context.

Warren: The implication there is that we need to be very careful about attributing any performance we see in Brazil. If this portfolio works, is it because the fundamental screen is brilliant, or because it’s simply picking stocks that have high exposure to things like value or size?

Iris: That’s what they are trying to isolate. They use a two-layer methodology precisely so they can decompose the returns into these systematic factors and genuine alpha generation. It sets up a test for whether the active strategy adds anything beyond what the market already prices in.

The paper's summary: Ruth: So, what do we actually see when we look at their findings? The central conclusion is pretty clear: this hybrid portfolio generated a Jensen alpha of one point one one percent per month in the Fama-French three-factor model, which they found to be statistically significant at the five percent level <ref:2611.06919#pg1,Jensen alpha of 1.11% per month>.

Warren: So, one point one one percent per month is what they are claiming as genuine alpha <ref:2611.06919#pg1>. That’s a monthly figure, so that translates to over thirty percent annually if you compound it—that’s a pretty substantial number for an active strategy in this market.

Iris: But the paper goes deeper than just that one number. They also found significant exposure to the size factor, SMB, with a beta of zero point three eight zero eight at a very high significance level, and they found no statistical significance for the value factor, HML.

Ruth: That’s a key point for us to track. It suggests that while there is alpha in this portfolio, most of that return is explained by exposure to the size risk premium rather than the traditional value premium we often expect from a fundamentalist screen.

Warren: And they also ran it through a four-factor model including momentum, where the alpha drops slightly to zero point nine eight percent per month, but it stays significant at a nine percent level for that specific test <ref:2611.06919#pg1>. So momentum adds some extra staying power to the signal.

Iris: The paper is very clear that by testing against these established models and looking at those factor loadings, they are trying to pinpoint exactly where the returns come from—whether it’s replicable risk premia or something truly new, like managerial skill.

The paper's improvements: Ruth: Now, the authors don't just stop there. They suggest several ways we can improve this kind of return decomposition work. One big improvement they propose is building a better factor attribution system using a four-factor model that includes momentum as a robustness test for their findings.

Warren: So, they want to see if the alpha holds up when you add momentum in there. That makes sense; it tests if the finding isn't just specific to the combination of value and size factors they tested earlier.

Iris: They also suggest incorporating dynamic strategy regime detection using sub-period analysis to see how market efficiency changes over time, which aligns with the idea from the adaptive markets hypothesis.

Ruth: That sounds like they are trying to account for changing market conditions, meaning their assumptions about what drives returns might shift depending on whether we’re in a developed or less developed regime.

Warren: And then there's this point about quantifying the risk cost of the technical timing component—how much "the moving average system introduces opportunity costs through frequent entries and exits and amplifies losses during sideways market movement."

Iris: That quantification is crucial because it lets us directly trade off that observed alpha of one point one one percent per month against the average drawdown they found, which was twenty-five point four nine percent for this portfolio.

Conclusion: Ruth: To wrap up, this paper on "Return Decomposition of Hybrid Strategies in the Brazilian Market: Evidence of Genuine Alpha and Factor Exposure" shows that when you combine a fundamental screen with technical timing in Brazil, you can find statistically significant alpha.

Warren: But we have to remember those numbers: one point one one percent per month in the three-factor model, but it’s heavily weighted toward the size factor exposure, SMB at zero point three eight zero eight.

Iris: The main implication is that asset selection strategies in this market often load up on known risk premia like size and value, so identifying genuine alpha beyond those factors is what’s fundamentally important for evaluating active management skill there.

Ruth: So the paper gives us a baseline, showing that while there's a positive return signal, we need to scrutinize what fraction of that return is actually from new insight versus just being exposed to known risk premiums.

Warren: That’s the trade-off they highlighted—the one point one one percent alpha versus that substantial twenty-five point four nine percent average drawdown for this hybrid strategy.

Iris: Ultimately, it provides a solid empirical test of how these hybrid models perform against standard factor models in an emerging market context like Brazil, setting the stage for future work on better risk cost quantification and regime detection.

Universidade Federal do Rio Grande do Norte

q-fin.ST, q-fin.PM

Submitted: 2026-05-09

Updated: 2026-05-09

Comments: 300 monthly observations (Jan/2000-Dec/2025). CAPM, Fama-French 3F and Carhart 4F models. OLS with robust standard errors. Jensen alpha: 1.11%/month (p=0.005). HML: p=0.544 (ns); SMB: p=0.000. Market beta approx 0.58. Sharpe difference vs Ibovespa not significant (Jobson-Korkie Z=1.427, p=0.154). Python scripts available

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 83/100

The gist: This paper decomposes the return sources of a hybrid portfolio combining a fundamentalist screen with a moving-average crossover system using simple moving averages to evaluate genuine alpha and…

Key concepts

Fundamentalist Screen
A strict set of financial rules used to select stocks based on metrics like low P/E ratios (under 15), high return on equity (over 15%), and strong margins. This method focuses on selecting fundamentally sound, high-quality companies rather than just market trends.
Moving Average Crossover System
A technical trading rule that uses two simple moving averages—one short-term (9 periods) and one long-term (46 periods)—to generate buy or sell signals. This system is used to time the entry and exit points for the selected stocks.
Jensen's Alpha
The excess return an investment generates above what is predicted by a financial model like the Fama-French three-factor model. A positive alpha indicates genuine, skill-based returns that are not explained by standard market risk factors.
SMB Factor (Small Minus Big)
One of the Fama-French factors representing the tendency for smaller companies to outperform larger ones over time. The study found this factor was highly significant in explaining a portion of the portfolio's positive returns.

Terminology

Summary

This paper decomposes the return sources of a hybrid portfolio combining a fundamentalist screen with a moving-average crossover system using simple moving averages to evaluate genuine alpha and factor exposure in the Brazilian market The central conclusion is unequivocal: the portfolio generated a Jensen alpha of 1. p. 0

Where the paper says it

br ABSTRACT This paper decomposes the return sources of a hybrid portfolio combining a fundamentalist screen — based on P/E (price-to-earnings), ROE (return on equity), Net Margin, Current Ratio, and P/BV (price-to-book value) — with a moving-average crossover (MAC) system using simple moving averages (SMA) of 9 and 46 periods, applied to the Brazilian stock exchange (B3) from January 2000 to December 2025 (300 monthly observations). CAPM and Fama-French three- and fourfactor models estimated by OLS (ordinary least squares) with heteroskedasticity-robust standard errors identify a positive and statistically significant Jensen alpha of 1.

11% per month (p = 0. p. 0
Where the paper says it

CAPM and Fama-French three- and fourfactor models estimated by OLS (ordinary least squares) with heteroskedasticity-robust standard errors identify a positive and statistically significant Jensen alpha of 1. 11% per month (p = 0.

p. 0
Where the paper says it

CAPM and Fama-French three- and fourfactor models estimated by OLS (ordinary least squares) with heteroskedasticity-robust standard errors identify a positive and statistically significant Jensen alpha of 1. 11% per month (p = 0.

005) in the Fama-French three-factor model

Portfolio Construction and Methodology

The study employs a two-layer methodology to investigate return decomposition In the first layer, a portfolio was built through a rigorous fundamentalist screen using ex-ante available data The selection parameters included (i) P/E ratio ≤ 15, (ii) ROE > 15%, (iii) Net Margin > 15%, (iv) Current Ratio ≥ 1, and (v) P/BV between 0.5 and 1.5 This screen was applied once at the beginning of the analysis period, generating a fixed set of 13 assets without periodic rebalancing The operationalization involved a trend-tracking algorithm based on a simple moving average crossover system with SMA of 9 and 46 periods

Factor Exposure Analysis

The econometric models were used to decompose return sources and test for alpha generation The CAPM model showed a statistically significant alpha of 1.18% per month (p = 0. p. 0

Where the paper says it

CAPM and Fama-French three- and fourfactor models estimated by OLS (ordinary least squares) with heteroskedasticity-robust standard errors identify a positive and statistically significant Jensen alpha of 1. 11% per month (p = 0.

016) The Fama-French three-factor model also revealed a positive and statistically significant Jensen's alpha of 1.11% per month (p = 0.005) The SMB factor was highly significant with a coefficient of βs = 0. p. 0
Where the paper says it

544): despite explicit value criteria in the screen, returns do not load on the value premium, likely because ROE and Net Margin requirements tilt the selection toward quality firms whose portfolios historically exhibit low correlation with pure HML. By contrast, the SMB (small-minus-big) factor is highly significant (βs = 0.

3808 (p = 0. p. 7
Where the paper says it

The SMB factor is positive and highly significant (βs = 0. 3808; t = 4.

000) Conversely, the HML factor was not statistically significant (βh = 0. p. 7
Where the paper says it

This pattern is consistent with the findings documented for the Brazilian market by Málaga and Securato (2004) and Mussa et al. The most unexpected and analytically significant result is the statistical non-significance of the HML factor (βh = 0.

1266; p = 0. p. 0
Where the paper says it

The central — and unexpected — finding is the non-significance of the HML (high-minuslow) factor (βh = 0. 1266; p = 0.

544) The four-factor model with momentum confirmed the robustness of the finding, showing an alpha of 0. p. 1
Where the paper says it

005) — evidence that challenges the semi-strong efficiency hypothesis in the Brazilian market. The four-factor model with momentum confirms the robustness of the finding: alpha of 0.

98% per month (p = 0. p. 0
Where the paper says it

CAPM and Fama-French three- and fourfactor models estimated by OLS (ordinary least squares) with heteroskedasticity-robust standard errors identify a positive and statistically significant Jensen alpha of 1. 11% per month (p = 0.

009)

Performance and Risk Assessment

Descriptive metrics indicated that the hybrid portfolio posted an average net profit of BRL 10.2 million over 25 years, exceeding the benchmark average The Sharpe ratio for the hybrid portfolio was 0. p. 8

Where the paper says it

2 Descriptive Portfolio Performance The fundamentalist screen demonstrated effectiveness in selecting historically highperformance companies: the portfolio posted an average net profit of BRL 10. 2 million over 25 years, exceeding the benchmark average (BRL 8.

443 versus 0. p. 0
Where the paper says it

Although the descriptive Sharpe ratio of the portfolio (0. 443) exceeds that of the Ibovespa (0.

039 for the Ibovespa, although this difference was not statistically significant under the Jobson-Korkie test (p = 0. p. 0
Where the paper says it

443) exceeds that of the Ibovespa (0. 039), the difference is not statistically significant under the Jobson-Korkie test (Z = 1.

154) The average drawdown of the hybrid portfolio was 25. p. 5
Where the paper says it

3 Descriptive Risk-Return Metrics The descriptive metrics of the first analytical layer include: (i) Total Net Profit; (ii) Maximum Drawdown — the largest percentage loss from peak to trough; (iii) Sharpe ratio (Sharpe, 1966); (iv) Sortino ratio (Sortino; Van der Meer, 1991); and (v) Profit Factor — the ratio of total gross profit to total gross loss across all operations. All metrics are calculated for the hybrid portfolio, for the buy-and-hold strategy on the same assets, and for the Ibovespa as the market benchmark.

49%, which is substantially higher than the benchmark's 10.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved Factor Attribution System: The AI can now decompose portfolio returns into genuine alpha, SMB factor exposure, and HML factor exposure using a four-factor model with momentum as a robustness test. This allows for precise quantification of which return sources are replicable by systematic risk premia versus those representing genuine managerial skill.

  2. Dynamic Strategy Regime Detection: The system can incorporate sub-period analysis to detect changes in market efficiency, aligning with the Adaptive Markets Hypothesis, by monitoring when positive alpha tends to concentrate in earlier sub-periods of the sample. This enables the AI to adjust its strategy assumptions based on whether it is operating in a more developed or less developed market regime.

  3. Risk Cost Quantification: The AI can calculate and explicitly present the concrete cost of the technical timing component, specifically quantifying how much the moving average system introduces opportunity costs through frequent entries and exits (whipsaws) and amplifies losses during periods of sideways market movement. This provides a direct trade-off analysis between the observed alpha (1.11% p.m.) and the associated average drawdown (25.49%).

  4. Quality Profile Assessment: The system can evaluate if the selected fundamentalist screen is primarily driven by value or quality criteria by testing for non-significance of HML when ROE and Net Margin requirements are included, as this suggests Dominance of quality over value in the portfolio construction. This allows for a more nuanced understanding of why returns do not load on pure HML despite explicit value criteria.

  5. Performance Fee Justification Module: The AI can generate a transparent report detailing what fraction of the return is genuine alpha versus replicable factor exposure, addressing the need to demonstrate to the investor what fraction of the return is genuine alpha, what fraction is SMB exposure (partially replicable), and what fraction is compensation for the additional drawdown.

Abstract

This paper decomposes the return sources of a hybrid portfolio combining a fundamentalist screen - based on P/E, ROE, Net Margin, Current Ratio, and P/BV - with a moving-average crossover system (SMA 9/46), applied to the Brazilian stock exchange (B3) from January 2000 to December 2025 (300 monthly observations). CAPM and Fama-French three- and four-factor models estimated by OLS with heteroskedasticity-robust standard errors identify a positive and statistically significant Jensen alpha of 1.11% per month (p = 0.005) in the three-factor model. In the four-factor model with momentum, alpha remains significant (0.98% p.m.; p = 0.009), while momentum is not significant (p = 0.409). The central - and unexpected - finding is the non-significance of the HML factor (beta h = 0.1266; p = 0.544): despite explicit value criteria, returns do not load on the value premium, likely because ROE and Net Margin tilt selection toward quality firms. By contrast, the SMB factor is highly significant (beta s = 0.38; p = 0.000), and the portfolio displays a defensive profile (market beta approximately 0.58). Although the descriptive Sharpe ratio of the portfolio (0.443) exceeds that of the Ibovespa (0.039), the difference is not statistically significant under the Jobson-Korkie test (Z = 1.427; p = 0.154). These findings advance the debate on semi-strong efficiency in the Brazilian market and carry direct implications for active strategy pricing and disclosure.

Related papers