Abstract

This study investigates how investor responses to earnings surprises vary systematically by sectoral classification and macroeconomic regime, with a focus on the 2007–2009 financial crisis as a natural stress test. Using Fama-French and Fama-MacBeth asset pricing frameworks, I show that cyclical sectors (e.g., Financials, Industrials) exhibit amplified reactions to earnings surprises during crisis periods, while defensive sectors (e.g., Healthcare, Consumer Staples) demonstrate more muted but persistent pricing behavior. Importantly, I find that market beta has limited predictive power in stable periods, but becomes significant only during crises, indicating that systematic risk is conditionally priced.

Through long-short portfolio strategies built on EPS surprise rankings, the analysis reveals that conditional alpha is most pronounced in resilient sectors during downturns—offering actionable insights for investors navigating macroeconomic stress. These findings highlight that post-earnings drift (PEAD) is not a uniform phenomenon; it is shaped by where a firm sits in the economy and when the signal arrives.

Empirically, this study draws on a panel of U.S. firms from 2003 to 2013, incorporating GICS sector classifications and macroeconomic segmentation (pre-crisis, crisis, post-crisis). By integrating cross-sectional and time-series techniques, the research contributes to a deeper understanding of how earnings information is processed under uncertainty, and offers a framework for sector-aware, crisis-conditioned investment strategies.

Key Takeaway: Earnings surprises are conditionally priced—underweighted in calm markets, and sharply magnified during crises, especially in cyclical sectors. Defensive sectors show weaker responses but lower volatility. In today’s environment of inflation, tariffs, and macro shocks, these historical patterns have returned—making this framework directly relevant to current market strategy.

1 Introduction

In financial markets, earnings surprises are not interpreted in a vacuum. A strong beat from Amazon in early 2009 sent its stock soaring, while a similar upside surprise from Coca-Cola barely moved the needle. Both exceeded expectations—but investor reactions diverged sharply. Why? Because markets don’t just respond to earnings numbers; they interpret them through the lens of sectoral risk, macroeconomic stress, and investor attention. This capstone examines the conditional pricing of earnings surprises—how their impact on stock returns shifts across industries and economic regimes. Drawing on cross-sectional and time-series methodologies, I show that the market’s reaction to earnings news is not uniform, but shaped by when the signal occurs and where the firm sits within the economic landscape. In crisis periods, surprises in cyclical sectors are amplified, while those in defensive industries often fade into noise—revealing a deeper structure to post-earnings drift.

Earnings surprises—deviations from consensus analyst expectations—are among the most potent catalysts for short-term stock price movement. Yet the broader pricing literature often treats firms homogeneously, failing to account for how sectoral positioning and macro stress distort these reactions. This study addresses that gap. Using the Fama-French and Fama-MacBeth frameworks, I quantify how cyclical and defensive sectors differ in their return responses to surprises, and how these effects intensify or attenuate during crisis periods. The result is a richer, more granular understanding of how investors price information when uncertainty is highest.

The 2007–2009 financial crisis provides a natural laboratory for these questions. Systemic shocks—Lehman’s collapse, liquidity freezes, cascading defaults—produced not just market-wide volatility, but stark sectoral divergence. Financials and real estate bore disproportionate losses, while healthcare and staples weathered the storm. Even within sectors, firms with positive earnings surprises were not uniformly rewarded—investor attention became selective, favoring resilience over risk. By capturing these dynamics through crisis-aware asset pricing models, this capstone contributes to both empirical finance and the practical design of investment strategies under stress.

The graph below compares the actual U.S. unemployment rate (2007–2012) with various Greenbook forecasts made by the Federal Reserve at different points in time. From this, we see the failure of economic forecasts to anticipate the severity of the 2008 financial crisis and its impact on unemployment. Initial projections significantly underestimated job losses, with even the worst-case stress scenario (Aug 2008) failing to capture the true scale of the downturn. The steep rise in unemployment following Lehman Brothers’ collapse exposed the limitations of traditional forecasting models in extreme economic shocks. The repeated upward revisions in projections reflect the Fed’s struggle to keep pace with rapidly deteriorating conditions, underscoring the challenge policymakers face in crisis management. The slow recovery post-2010 further emphasizes how deep recessions create prolonged labor market scarring, reinforcing the need for aggressive and adaptive policy responses in times of financial turmoil.

The next graph reveals striking parallels between 2008 and the Great Depression, particularly in terms of stock market declines, house prices, and household wealth. While the Great Depression unfolded over a longer period, the initial shocks of the 2008 crisis were more severe, with the stock market plunging 57.8% compared to 44.9% in the early stages of the Great Depression. Likewise, house prices fell three times as much in 2008 (-18.3%) as they did during the Great Depression (-6.2%), amplifying the financial strain on homeowners and contributing to widespread economic distress. The steeper decline in household wealth (-14.1%) further illustrates the increasing reliance on financial assets in modern economies, making individuals more vulnerable to rapid market fluctuations. These patterns reinforce the heightened fragility of financialized economies and the necessity for swift, large-scale policy interventions to contain systemic risks and prevent prolonged downturns.

This capstone addresses a critical gap in asset pricing research: how sectoral characteristics condition the market’s response to earnings surprises—especially during periods of economic stress. Leveraging the Global Industry Classification Standard (GICS), I systematically evaluate how different sectors absorb and reflect earnings information across distinct macroeconomic regimes. By embedding these insights within robust empirical frameworks—including the Fama-French three-factor model and the Fama-MacBeth regression methodology—this study delivers a granular, time-aware analysis of how risk, valuation, and information shocks interact across industries. These models not only capture the explanatory power of market-wide factors like size and value, but also reveal how their influence shifts across sectors and business cycles.

While the empirical focus spans 2003 to 2013, the relevance of this research is decidedly contemporary. From the COVID-19 pandemic to renewed geopolitical instability, today’s markets continue to be shaped by macro-level disruptions. The earnings shock playbook remains sector-specific: during COVID, e-commerce firms like Amazon soared on positive surprises, while travel and hospitality languished. In the current climate of sticky inflation, policy volatility, and protectionist shocks, understanding how industries price earnings surprises under stress is more essential than ever. This project equips investors and policymakers with a framework to decode these patterns—helping navigate today’s fractured markets with sharper tools and more targeted strategies.

This is not just a historical curiosity. The same dynamics have returned with a vengeance in early 2025. Under a renewed Trump administration, aggressive tariffs, fiscal volatility, and inflation have roiled global markets. Investors are once again fleeing from cyclicals and tech, rotating into traditional safe havens like Healthcare and Utilities. Earnings surprises in these resilient sectors are being priced more sharply than ever, while those in volatile industries often get ignored or punished. The strategies explored in this capstone—long-short earnings plays, sector-aware pricing, and beta/surprise interaction—mirror exactly what’s unfolding in real time. As an example, we observe the two graphs below, dated April 28th, 2025, courtesy of Apollo chief economist Torsten Slok.

These two charts underscore the macro undercurrents driving today’s market behavior. The surge in households making only minimum credit card payments signals rising financial strain, which directly undermines discretionary spending and cyclical sector earnings. Simultaneously, collapsing consumer sentiment—across all income levels—adds fuel to the rotation into defensive sectors. Together, these trends amplify the pricing of earnings surprises in Healthcare, Utilities, and other low-beta sectors, precisely validating the crisis-contingent pricing dynamics modeled in this study.

1.1 Study Contributions

This study enhances the earnings surprise literature by embedding sectoral identity and macro regime context into asset pricing models. Its key contributions are:

Sector × Crisis Interactions:
By combining GICS classifications with macroeconomic segmentation (Pre-Crisis, Crisis, Post-Crisis), the study uncovers how earnings surprises are priced differently across industries and stress periods—exposing patterns that homogeneous models miss.
Fama-MacBeth with Conditional Risk Premia:
Using Fama-MacBeth regressions on firm-level data, the study estimates time-varying risk premiums on both market beta and EPS surprise, showing that these factors gain or lose explanatory power depending on the macro environment.
Stress-Tested PEAD Strategies:
The analysis tests long-short earnings surprise strategies across sector and crisis combinations, revealing that post-earnings drift (PEAD) is strongest in defensive sectors during downturns—validating the conditional nature of alpha.

1.2 Research Questions & Hypotheses

How do GICS industries differ in their return responses to earnings surprises across market regimes?
Did the 2007–2009 crisis amplify sector-specific reactions to earnings surprises?
How do long-short strategies built on surprise signals perform across sectors and stress conditions?

H1: Sectoral Heterogeneity
Cyclical sectors react more strongly to earnings surprises than defensive ones.

H2: Crisis Amplification
Earnings surprises have larger price impacts during crisis periods due to volatility and attention shifts.

H3: Conditional Strategy Alpha
Earnings-based long-short portfolios earn higher returns in resilient sectors under macro stress.

2 Literature Review

The relationship between earnings surprises, sectoral dynamics, and financial crises has been a long-standing area of interest in finance. While significant strides have been made in understanding the impacts of earnings surprises on stock returns, the interplay between these factors remains underexplored. This review synthesizes key findings across these dimensions, focusing on their relevance to the proposed study.

2.1 Earnings Surprises and Market Reactions

Earnings surprises have been studied extensively as a driver of stock price movements. The foundational study by Ball and Brown (1968) demonstrated that stock prices react significantly to unexpected earnings announcements, with both positive and negative surprises resulting in predictable price adjustments. This seminal work was extended by Bernard and Thomas (1990), who documented the post-earnings announcement drift (PEAD), a phenomenon where markets underreact to earnings surprises, leading to abnormal returns over time.

Further research has explored the behavioral underpinnings of earnings surprise effects. Barberis et al. (1998) argued that cognitive biases, such as representativeness and overconfidence, delay market responses to earnings announcements. Similarly, Doyle et al. (2006) showed that extreme earnings surprises often lead to long-term outperformance, highlighting the persistence of these market inefficiencies. However, these studies largely focused on individual stocks and did not delve into sector-specific patterns.

The role of investor expectations and market sentiment in shaping reactions to earnings surprises has also been a subject of inquiry. Keung et al. (2010) found that small positive earnings surprises often generate muted market reactions compared to larger surprises. Meanwhile, Skinner and Sloan (2002) demonstrated that growth stocks overreact more strongly to negative surprises than value stocks, suggesting that firm characteristics influence the magnitude of the reaction.

2.2 Sectoral Differences in Earnings Surprise Dynamics

Sectoral dynamics play a critical role in shaping the impact of earnings surprises. Industries differ in their sensitivity to earnings announcements due to variations in operational risk, competitive structure, and investor base. For instance, Brown et al. (2008) found that positive earnings surprises in high-growth sectors reduce information asymmetry, while negative surprises amplify it. However, their work stopped short of examining how these dynamics manifest during periods of economic stress.

In the context of risk-adjusted returns, Fama and French (1993) introduced the three-factor model, which highlights the influence of market excess returns, size (SMB), and value (HML) factors. While this model has been extensively validated, its application to sectoral analysis remains limited. Sector-specific studies, such as Abarbanell and Park (2016), suggested that small earnings surprises have minimal market impact after controlling for behavioral biases. Yet, this research did not address portfolio-level effects or crisis-period dynamics.

More recent studies have sought to bridge this gap. For example, Chen and Zhang (2018) explored how technology and healthcare stocks exhibit more pronounced reactions to earnings surprises due to their growth potential and investor attention. In contrast, defensive sectors like utilities and consumer staples showed muted responses, highlighting the heterogeneity in sectoral behavior.

2.3 Earnings Surprises During Financial Crises

The financial crisis of 2007–2009 provides a unique context for studying earnings surprises. During this period, market inefficiencies were exacerbated by liquidity shocks and heightened uncertainty. Brunnermeier (2009) demonstrated how liquidity constraints amplified pricing inefficiencies, including the mispricing of earnings surprises. Similarly, Gorton (2008) argued that the opacity of financial instruments during the crisis led to irrational market behavior, further distorting the relationship between earnings announcements and stock returns.

Bharath and Shumway (2008) explored how the heightened risk aversion during the crisis led to stronger reactions to negative earnings surprises, as investors prioritized downside protection. Meanwhile, Fahlenbrach and Stulz (2011) analyzed how firm-specific characteristics, such as leverage and exposure to subprime assets, influenced earnings surprise impacts, particularly in the financial sector. Behavioral studies have also shed light on crisis-period dynamics. Barberis et al. (2008) linked behavioral biases, such as loss aversion and herding, to the exaggerated market reactions observed during the crisis. These findings suggest that earnings surprises during crises are not only a reflection of firm performance but also a product of broader market psychology.

2.4 Integration of Earnings Surprises, Sectoral Dynamics, and Crises

While significant progress has been made in understanding earnings surprises and market reactions, few studies have integrated sectoral dynamics and crisis-period analysis. For example, Dichev and Piotroski (2001) found that distressed firms exhibited more volatile reactions to earnings surprises, but their work did not examine how this volatility varied across industries or during crises. Similarly, Campbell et al. (2008) documented that macroeconomic shocks influence earnings surprise impacts, yet their analysis was limited to aggregate market behavior. Recent advances in asset pricing models, such as the integration of GICS industry codes into factor models, provide a promising avenue for addressing these gaps. By analyzing earnings surprises at the sectoral level and across different market conditions, researchers can uncover the nuanced interplay between firm-specific and macroeconomic factors.

3 Theoretical Framework

Understanding how markets price earnings surprises—especially under stress—requires grounding in three interlocking strands of theory: asset pricing models, market efficiency and behavioral finance, and crisis economics. Together, these lenses explain not only what investors price, but when and why certain information gains salience. This section outlines the theoretical foundations that guide our empirical inquiry.

3.1 Asset Pricing Theory: Factor Models and Conditional Risk

At the core of this study is the Fama-French three-factor model, a widely accepted extension of the traditional Capital Asset Pricing Model (CAPM). This model posits that expected stock returns are explained not only by exposure to market-wide risk, but also by firm size and value orientation. The three factors are:

Market excess return (MKT): Return of the market portfolio over the risk-free rate.
Size (SMB – Small Minus Big): The return differential between small-cap and large-cap stocks.
Value (HML – High Minus Low): The return differential between high book-to-market (value) and low book-to-market (growth) stocks.

These factors proxy for systematic risks that traditional CAPM omits. Crucially for this project, factor loadings may not be stable across time or sectors. For instance, cyclical industries tend to exhibit higher market betas and stronger loading on value factors. By incorporating these dynamics, the model allows us to estimate how sector identity shapes exposure to priced risk factors, particularly in volatile regimes.

Additionally, the study employs Fama-MacBeth cross-sectional regressions to capture time-varying prices of risk. These regressions explicitly allow risk premia to shift across macroeconomic environments, revealing how investor preferences evolve under stress. The result is a dynamic rather than static view of risk-return tradeoffs—essential when exploring crisis-period behavior.

3.2 Market Efficiency vs. Behavioral Finance: Information Delays and Drift

Under the Efficient Market Hypothesis (EMH), stock prices instantly reflect all available information, including earnings announcements. But empirical anomalies—most notably, post-earnings announcement drift (PEAD)—challenge this view. Numerous studies show that stock prices underreact to earnings surprises, generating abnormal returns over subsequent months.

Behavioral finance offers explanations for these inefficiencies. Models of investor underreaction, attention constraints, and limited arbitrage suggest that information is not always absorbed immediately—particularly in noisy, distracted, or crisis environments. Moreover, overreaction to salient surprises, especially in volatile industries, can lead to subsequent reversals.

Importantly, these behavioral biases are often sector-specific. For instance, investors may be more attentive to tech earnings than to those from utilities. This study builds on these insights by exploring how investor reactions to surprises differ across GICS sectors—and how attention and signal-processing behavior shifts under macro stress.

3.3 Crisis Economics: Information Amplification Under Stress

The 2007–2009 financial crisis—and crises more broadly—offer a unique lens into pricing behavior. In periods of panic, standard models often break down: liquidity dries up, risk aversion spikes, and cross-asset correlations increase. But crises also serve as magnifying glasses: they reveal what markets really care about when noise is stripped away.

Theoretical work by Brunnermeier (2009) and Gorton (2008) shows that during crises, information asymmetries worsen, behavioral biases intensify, and investor focus shifts dramatically toward firm resilience and systemic exposure. As a result, the same earnings surprise may generate radically different price reactions in stable vs. stressed environments. This gives rise to conditional pricing, where the value of information depends on context.

This study leverages crisis periods to examine how systematic risk and firm-specific news interact. The core hypothesis: Earnings surprises are more heavily priced when uncertainty is highest—and sectoral identity determines how that signal is processed.

3.4 Synthesis

By integrating factor-based asset pricing, behavioral inefficiencies, and crisis theory, this framework enables a nuanced, conditional analysis of how earnings surprises are priced. It treats pricing not as a static equation but as a moving target—shaped by the intersection of sector risk, investor behavior, and macro regime. This triangulation sets the stage for the empirical sections that follow, which test these interactions using CAPM, Fama-French, and Fama-MacBeth methodologies across sectoral and crisis-period breakdowns.

4 Methodology

This study uses a structured quantitative approach to explore how stock returns react to earnings surprises across sectors and economic regimes. The methodology combines portfolio sorting, asset pricing models, and cross-sectional regressions to capture both time-series and cross-sectional dynamics.

Stock returns, earnings data, and analyst forecasts were sourced from CRSP, Compustat, and I/B/E/S.
Earnings surprises were calculated as:
\[ \text{EPS Surprise} = \text{Actual EPS} - \text{Consensus Forecast} \]
Firms were categorized using GICS sector codes to enable sector-level comparisons, especially between cyclical and defensive industries.
The sample period spans 2003 to 2013, segmented into:
- Pre-Crisis: 2003–2006
- Crisis: 2007–2009
- Post-Crisis: 2010–2013
Macroeconomic variables (e.g., market excess returns, risk-free rates) were obtained from FRED and Fama-French data libraries.

Each quarter, firms were ranked by earnings surprise, and sorted into:
- Top 20 Portfolio: firms with the highest positive surprises
- Bottom 20 Portfolio: firms with the lowest (or most negative) surprises
- Long-Short Strategy: long Top 20, short Bottom 20
Monthly excess returns were calculated for each strategy.
Cumulative return plots were created to visualize how each portfolio performs over time and across economic regimes.

Predictive regressions were estimated to test whether lagged earnings surprises predict excess returns: \[ \text{Excess Return}_{i,t} = \alpha + \beta \cdot \text{EPS Surprise}_{i,t-k} + \epsilon_{i,t} \]
- Regressions were run using 1-year and 5-year lags.
- Metrics reported: coefficients, R², standard error ratios (conditional vs. unconditional volatility).

To assess exposure to systematic risk, CAPM and Fama-French 3-Factor models were estimated:
- CAPM: Market excess return only
- FF3:
  \[ \text{Excess Return}_{i,t} = \alpha + \beta_1 \cdot \text{MKT} + \beta_2 \cdot \text{SMB} + \beta_3 \cdot \text{HML} + \epsilon_{i,t} \]
- Regressions were run for:
  - Top 20 Portfolio
  - Bottom 20 Portfolio
  - Long-Short Strategy

Fama-MacBeth two-step regressions were used to estimate time-varying risk premia:
- Step 1: Estimate each firm’s beta via time-series regression on market returns
- Step 2: Each month, regress firm excess returns on beta and EPS surprise: \[ \text{Excess Return}_{i,t} = \lambda_1 \cdot \hat{\beta}_{i} + \lambda_2 \cdot \text{EPS Surprise}_{i,t} + \epsilon_{i,t} \]
- Regressions were run across:
  - All firms (full sample)
  - By GICS sector
  - By crisis period
  - By sector × crisis period combinations

Visualizations and diagnostics include:
- Cumulative return plots for all three portfolio strategies
- Heatmaps of observation counts by sector and crisis period
- Histograms of estimated λ₂ (beta) and λ₃ (EPS surprise) premiums
- T-tests and confidence intervals for average λ values
- Variance and standard errors of λ coefficients to assess pricing consistency

This comprehensive design enables the study to answer when and where earnings surprises are most predictive of returns—highlighting that their impact is not uniform, but conditional on sectoral identity and macroeconomic stress.

5 Predictive Regressions

Predictive in-sample regressions are a vital statistical tool used to examine the relationship between variables within the same dataset from which the model is developed. In our analysis, these regressions test the hypothesis that earnings surprises significantly affect stock returns. By using historical data, we are able to construct and estimate regression models that evaluate how past earnings surprises relate to the performance of stock returns within the same period.

5.1 One-Year EPS Surprise Lag

First, we test our hypothesis using a one-year EPS Surprise Lag. Therefore, we expect that a one-year lag introduces sufficient time for returns to digest positive or negative EPS Surprises. Thus, the excess returns of our portfolio is regressed against the EPS Surprise of our portfolio for one year prior, with the results as shown below:

***Regression Results***

	Portfolio Excess Return Over Rf

EPS Surprise - 1 Year Lag	0.004
	(0.007)
Constant	0.009
	(0.006)
N	115
R²	0.003
Adjusted R²	-0.006
Residual Std. Error	0.055 (df = 113)
F Statistic	0.376 (df = 1; 113)

p < .1; p < .05; p < .01

We can see that the coefficient of 0.004 suggests that for each unit increase in earnings surprise, the excess return over the risk-free rate is expected to increase by 0.004 units. However, this effect is not statistically significant, as indicated by the standard error of 0.007, which is larger than the coefficient itself. The constant of 0.009 implies that when the EPS surprise is zero, the excess return over the risk-free rate is expected to be 0.009. This also lacks statistical significance due to the standard error of 0.006 being close in magnitude to the coefficient. Additionally, The R² value of 0.003 is extremely low, indicating that only 0.3% of the variance in the portfolio’s excess returns is explained by this model. This highlights a very weak explanatory power.

The calculated ratio of the standard deviation of predicted returns to the standard deviation of observed returns, for a one-year lag, is 0.057592. This ratio is a measure of the predictive power of the regression model, and indicates here that the predicted returns from the regression model explain only a small portion of the variability in the actual observed returns.

A ratio as low as 0.057592 suggests that other factors not included in the model might be influencing the returns more significantly. It could indicate the need for additional variables to better capture the complexity of stock price movements or that the dynamics of stock returns are influenced by a myriad of other market, economic, or firm-specific factors not accounted for by earnings surprises alone.

5.2 Five-Year EPS Surprise Lag

We run this regression again, but now with a five-year lag in the EPS returns. The five-year lag period is particularly chosen to assess the enduring effects of earnings surprises, beyond the immediate market reactions typically captured in shorter time frames.

For the five-year EPS Surprise lag in the predictive in-sample regression, the calculated ratio of the standard deviation of predicted returns to the standard deviation of observed returns is 0.1126859. The ratio of 0.1126859 is relatively low, indicating that the regression model’s predicted returns based on the EPS Surprise five years prior explain only a small fraction of the volatility observed in actual returns. This suggests that the model, with EPS Surprise as a predictor, has limited effectiveness in capturing the full dynamics of excess returns over the risk-free rate.

The regression results for this are as shown below:

***Regression Results***

	Portfolio Excess Return Over Rf

EPS Surprise - 5 Year Lag	-0.110
	(0.212)
Constant	-0.002
	(0.013)
N	23
R²	0.013
Adjusted R²	-0.034
Residual Std. Error	0.029 (df = 21)
F Statistic	0.270 (df = 1; 21)

p < .1; p < .05; p < .01

The coefficient for this regression is -0.110 with a high standard error (0.212), suggesting that a positive earnings surprise five years prior might slightly decrease excess returns, but this effect is not statistically significant. The F-statistic of 0.270 and an expected high p-value indicate that the results are not statistically significant, implying that the model fails to provide convincing evidence against the null hypothesis of no effect. Additionally, R-squared is extremely low (0.013), indicating that the model explains only 1.3% of the variability in the excess returns.

6 The Fama-French Method

The Fama-French three-factor model is a well-regarded framework in financial economics that expands on the traditional Capital Asset Pricing Model (CAPM) by incorporating two additional factors—size and value—into the analysis of stock returns. Developed by Eugene Fama and Kenneth French, this model acknowledges that beyond market risk, the size of firms (small versus large) and their book-to-market values (high versus low) significantly contribute to explaining the variability in stock returns across different portfolios. The model is particularly valued for its ability to explain historical variations in portfolio returns more accurately than CAPM.

In our study, we employ the Fama-French model to examine how well these three factors—market risk, size, and value—explain the returns of portfolios that are sorted based on their sensitivity to earnings surprises. This approach not only provides a deeper understanding of the factors that drive returns but also tests the robustness of earnings surprises as a predictor of stock performance under the Fama-French framework.

To implement this, we regress the excess returns of these portfolios against the Fama-French three factors:

Market Excess Return (MKT): The excess return on the market portfolio over the risk-free rate.
Size (SMB, Small Minus Big): The excess return of portfolios with small capitalizations over those with large capitalizations.
Value (HML, High Minus Low): The excess return of portfolios with high book-to-market ratios over those with low book-to-market ratios.

Using EPS Surprise, we sort our universe of 100 companies each quarter, into two distinct portfolios - our top portfolio, consisting of 20 companies with the highest positive EPS Surprise from last quarter, and our bottom portfolio, consisting of 20 companies with the lowest positive EPS Surprise from last quarter. This sorting is done every quarter over the estimation period of 10 years, and the cumulative returns made from the top portfolios, the bottom portfolios, and a long-short strategy consisting of both is plotted below.

6.1 Visualizing Portfolio Returns

Some initial volatility is observed, especially around 2008, where the returns fluctuate significantly. Post the initial volatile period, the portfolio shows a steady upward trend in cumulative returns starting from around 2009. This suggests that companies with high positive EPS surprises consistently performed well, likely due to investor confidence in these companies’ ability to exceed earnings expectations.From 2012 onwards, the growth in cumulative returns becomes more pronounced, indicating a period of sustained high performance. This could be attributed to the compounding effects of reinvesting earnings in a portfolio that consistently selects high-performing stocks based on their earnings surprises. As we can see in the portfolio, an investment of $1 made in this portfolio at the beginning of 2007 would have become a total of about $2.5 by beginning 2015.

The returns for our shorted portfolio for our bottom 20 portfolios are plotted below.

The returns for the long-short portfolio, where we go long on the top20 portfolios and short the bottom 20, are as below:

The returns for all our various strategies are plotted as below.

Thus, overall, the sharp movements during 2008-2009 across all strategies highlight the impact of the financial crisis and subsequent recovery. The long Top 20 strategy clearly outperformed the other strategies, especially from 2012 onwards, showing the benefit of selecting stocks with strong earnings surprises. The stabilization of the short Bottom 20 strategy post-crisis suggests a limit to the gains from shorting poor performers in a recovering or stable market.

The combined strategy’s performance suggests that while it can reduce risk by offsetting positions, it may also cap potential gains, leading to more stable but less dramatic returns compared to a pure long strategy.

This analysis shows the importance of market timing, selection criteria, and the broader economic environment’s impact on investment strategies. The effectiveness of each strategy can vary significantly depending on these factors, as vividly demonstrated by the varied trajectories of the three strategies through the years.

6.2 The Fama-French Method Regressions

We now run a total of six regressions, as follows:

Top 20 Portfolio vs. the Market Factor
Long-Short Portfolio vs. the Market Factor
Top-20 Portfolio vs. the Fama French 3 Factors
Bottom-20 Portfolio vs. the Fama French 3 Factors
Bottom-20 Portfolio vs. the Market Factor
Long-Short Portfolio vs. the Fama French 3 Factors

Through these regressions, we explore how well the market factor alone explains portfolio returns, as well as the 3 factors. We begin, therefore, with the Top 20 Portfolio vs. the Market Factor regression.

***Regression Results***

	Portfolio Excess Return Over Rf

Top 20 Portfolio - Market Factor	0.774^***
	(0.038)
Constant	0.003
	(0.002)
N	115
R²	0.786
Adjusted R²	0.784
Residual Std. Error	0.022 (df = 113)
F Statistic	414.085^*** (df = 1; 113)

p < .1; p < .05; p < .01

The coefficient for the Excess Return of Portfolio is 0.773809, suggesting a strong positive relationship with the Top20ReturnOverRf. This implies that for each unit increase in the excess return of the market portfolio, the top 20 returns sorted by earnings surprise increase by approximately 0.774 units.

The p-value associated with this coefficient is extremely small (<2e-16), indicating that the relationship is statistically significant and the likelihood of this result occurring by chance is very low. The intercept is 0.002836, which represents the expected value of Top20ReturnOverRf when the Excess Return of Portfolio is zero.

The intercept is not statistically significant (p-value = 0.182), suggesting that when the excess return of the market portfolio is zero, the top 20 returns do not significantly differ from zero. The R-squared value of 0.7856 means that approximately 78.56% of the variation in the top 20 returns is explained by the excess returns of the market portfolio.

Thus, the F-statistic of 414.1 with a p-value less than 2.2e-16 strongly rejects the null hypothesis that the excess return of the market portfolio has no effect on the top 20 returns.

***Regression Results***

	Portfolio Excess Return Over Rf

Long-Short Portfolio vs. Market Factor	-0.376^***
	(0.074)
Constant	0.004
	(0.004)
N	115
R²	0.186
Adjusted R²	0.178
Residual Std. Error	0.044 (df = 113)
F Statistic	25.763^*** (df = 1; 113)

p < .1; p < .05; p < .01

For the Long-Short Portfolio vs. Market Factor regression, our coefficient of -0.376216 indicates a negative relationship between the excess return of the market portfolio and the long-short excess returns. This suggests that as the market performs better, the long-short strategy yields lower excess returns, relative to the market. The relationship is statistically significant, with a p-value of approximately 1.53e-06, which is far below the conventional alpha level of 0.05.

About 18.57% of the variability in LongShortExcess is explained by the excess returns of the market portfolio. This is relatively low, indicating that other factors not included in the model might also be influencing the long-short excess returns.

***Regression Results***

	Portfolio Excess Return Over Rf

Top 20 Portfolio - 3-Factor	0.812^***
	(0.044)
SMB	0.028
	(0.100)
HML	-0.202^**
	(0.088)
Constant	0.002
	(0.002)
N	115
R²	0.796
Adjusted R²	0.790
Residual Std. Error	0.022 (df = 111)
F Statistic	143.953^*** (df = 3; 111)

p < .1; p < .05; p < .01

In our Top 20 Portfolio - 3-Factor regression, our estimate is 0.812329, a substantial positive coefficient indicating a strong relationship between excess market returns and the top 20 adjusted returns. This suggests that higher market returns significantly correlate with higher returns in the top 20 portfolio.

This estimate is extremely significant with a p-value of <2e-16, confirming the robustness of this predictor. The estimate for the size factor is 0.028444, suggesting a very small positive influence of small minus big cap stocks on the top 20 returns. However, this is not statistically significant (p-value = 0.7756), indicating that the size premium does not significantly influence the portfolio returns in this model. The value factor has an estimate of -0.202470, showing a negative relationship between the value premium and the top 20 returns. This is statistically significant (p-value = 0.0229), implying that higher value stock returns correlate with lower returns in the top 20 portfolio.

An r-squared of 0.7955 indicates that about 79.55% of the variation in the top 20 portfolio returns is explained by the model, which is a high degree of explanatory power.

***Regression Results***

	Portfolio Excess Return Over Rf

Bottom 20 Portfolio - 3-Factor	1.156^***
	(0.048)
SMB	-0.209^*
	(0.109)
HML	0.126
	(0.096)
Constant	-0.001
	(0.002)
N	115
R²	0.877
Adjusted R²	0.873
Residual Std. Error	0.024 (df = 111)
F Statistic	262.973^*** (df = 3; 111)

p < .1; p < .05; p < .01

For the short portfolio’s regression against the 3 factors, the estimate is 1.1559662, indicating a strong positive impact on the short portfolio returns. This implies that as market returns increase, so do the returns of the short portfolio, adjusted for risk-free rate. This is extremely significant with a p-value of <2e-16, suggesting a robust influence.

About 87.67% of the variability in the short portfolio returns is explained by the model, indicating excellent explanatory power.

***Regression Results***

	Portfolio Excess Return Over Rf

Bottom 20 Portfolio - Market Factor	1.153^***
	(0.042)
Constant	-0.001
	(0.002)
N	115
R²	0.871
Adjusted R²	0.870
Residual Std. Error	0.025 (df = 113)
F Statistic	761.684^*** (df = 1; 113)

p < .1; p < .05; p < .01

The regression results for the Bottom 20 Portfolio against the Market Factor show a highly significant relationship, indicating that the excess return of this portfolio over the risk-free rate is strongly predicted by market movements. The coefficient for the market factor is 1.153, with a very tight standard error of 0.042, suggesting a robust and statistically significant positive correlation. This coefficient implies that for every one percent increase in market returns, the bottom 20 portfolio’s excess return over the risk-free rate increases by about 1.153 percent, a substantial leverage effect.

The constant term is effectively zero (-0.001) with a standard error of 0.002, indicating that in the absence of market returns, the portfolio’s excess return is virtually zero and not significantly different from the risk-free rate. This suggests that the performance of the bottom 20 portfolio is heavily dependent on the market’s performance.

The R-squared value of 0.871 indicates that approximately 87.1% of the variation in the portfolio’s excess returns is explained by the market factor alone, which is extremely high.

***Regression Results***

	Portfolio Excess Return Over Rf

Long-Short Portfolio - Three Factors	-0.342^***
	(0.086)
SMB	0.240
	(0.194)
HML	-0.326^*
	(0.171)
Constant	0.002
	(0.004)
N	115
R²	0.221
Adjusted R²	0.200
Residual Std. Error	0.043 (df = 111)
F Statistic	10.508^*** (df = 3; 111)

p < .1; p < .05; p < .01

For the regression for the long-short portfolio against the three factors, we see a significant negative effect; more market excess return leads to lower long-short strategy returns. The SMB factor is not significant; there is no clear effect from the size premium.

The R-squared is 22.12%, showing modest explanation of returns by the model. Overall significance is very strong with a p-value of 3.87e-06. The model suggests market conditions significantly impact long-short strategy performance, but only a portion of the variability is explained by these factors.

The series of six regression analyses conducted provides valuable insights into how different portfolio strategies respond to market dynamics and specific financial factors. The regressions clearly showed that both the Top 20 and Bottom 20 portfolios are significantly influenced by market factors, with the Bottom 20 portfolio showing an especially high dependency on market movements, as evidenced by its high R-squared value. When the Fama-French three factors were introduced, they significantly impacted the analysis, particularly by explaining a large portion of the variance in returns for the Top 20 and Bottom 20 portfolios. This suggests that size and value factors, alongside market dynamics, play crucial roles in the performance of these portfolios. The Long-Short strategy results indicated a negative correlation with market performance, suggesting its potential utility as a hedge against market volatility. Overall, the strong statistical significance and robust explanatory power of these models underscore the critical influence of both market and macroeconomic factors on portfolio performance, highlighting the necessity of incorporating such factors into strategic portfolio management.

7 The Fama-Macbeth Method

The Fama-Macbeth method, developed by Eugene Fama and James MacBeth, is a robust approach to estimating the relationship between financial variables and asset returns across multiple time periods. This two-step regression method is widely used in empirical finance research, particularly for testing asset pricing models. The first step involves conducting time-series regressions to estimate betas (risk factors) for each asset.

These betas are then used in the second step, which involves cross-sectional regressions for each time period to estimate the price of risk and test the asset pricing model. The method’s strength lies in its ability to control for cross-sectional correlation in the error terms, providing more reliable and consistent estimates that are crucial for understanding and predicting asset returns. In this study, the Fama-Macbeth approach is employed to analyze the effects of market factors and specific financial variables, such as earnings surprises, on the returns of different portfolios, providing a comprehensive view of how these factors influence investment outcomes over time.

7.1 Obtaining Beta - Time Series Regressions

7.1.1 Sector and Crisis Period Integration

To extend our analysis across economic cycles and industry dynamics, we introduce two key categorical variables into our dataset: GICS Sector and Crisis Period. The Sector variable classifies each firm into its respective GICS sector (e.g., Financials, Healthcare), while the Crisis_Period variable segments the data temporally into Pre-Crisis (2003–2006), Crisis (2007–2009), and Post-Crisis (2010–2013) periods.

The snippet below displays a sample of the newly added columns to confirm their structure and presence.

Sample of Sector and Crisis_Period Columns
Company	Sector	Crisis_Period
ORCL	Information Technology	Pre-Crisis
ORCL	Information Technology	Pre-Crisis
ORCL	Information Technology	Pre-Crisis
ORCL	Information Technology	Pre-Crisis
ORCL	Information Technology	Pre-Crisis
ORCL	Information Technology	Pre-Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis
ORCL	Information Technology	Post-Crisis

We next summarize the distribution of firms across sectors and the number of observations across crisis periods. This helps validate that we have sufficient representation across the combinations and allows us to identify any imbalances in the panel.

To further explore the joint structure of our data, we create a cross-tabulation of observations by sector and crisis period. This matrix helps identify which sectors are heavily concentrated in particular economic periods, which is important context for our later regression analysis.

7.1.2 Visualizing Sector and Crisis Composition

To complement the tabular summaries, we use simple visualizations to better understand the structure of our dataset:

A bar chart showing the number of firm-month observations per GICS sector
A bar chart showing the number of observations per Crisis Period
A heatmap showing the interaction between sector and crisis period

These plots allow us to spot overrepresented or underrepresented sectors and periods, ensuring robustness in our econometric estimates.

The Beta’s that we obtained for our universe of companies are shown below, alongside their standard error and standard error as a % of Beta.

Estimated Betas for Companies
Company	Beta Estimate	Standard Error	Standard Error as %
ORCL	0.847	0.088	10.39
MSFT	0.779	0.100	12.84
HON	1.028	0.062	6.03
EMC	0.850	0.108	12.71
UIS	3.076	0.237	7.70
KO	0.401	0.068	16.96
DD	1.265	0.081	6.40
XOM	0.404	0.071	17.57
GD	0.956	0.072	7.53
GE	1.239	0.084	6.78
GM	1.039	0.225	21.66
IBM	0.564	0.076	13.48
HCA	1.005	0.255	25.37
CZR	1.439	0.925	64.28
PEP	0.384	0.061	15.89
MO	0.340	0.079	23.24
AMGN	0.453	0.114	25.17
S	0.890	0.265	29.78
KODK	1.294	0.634	49.00
SLB	1.037	0.121	11.67
SNSTA	0.791	0.281	35.52
CVX	0.625	0.082	13.12
ERI	0.861	0.619	71.89
KHC	0.481	0.377	78.38
TXN	0.887	0.088	9.92
UTX	0.871	0.056	6.43
PG	0.386	0.065	16.84
SO	0.159	0.062	38.99
CAT	1.428	0.104	7.28
CL	0.380	0.060	15.79
BMY	0.417	0.102	24.46
BA	0.975	0.094	9.64
BDK	1.398	0.201	14.38
ABT	0.360	0.081	22.50
DOW	1.884	0.134	7.11
IP	1.829	0.116	6.34
EXC	0.366	0.092	25.14
PFE	0.571	0.077	13.49
JNJ	0.450	0.055	12.22
MMM	0.740	0.066	8.92
MRK	0.498	0.097	19.48
SLE	0.662	0.111	16.77
HSH	1.163	0.863	74.20
HAL	1.204	0.141	11.71
ETR	0.361	0.088	24.38
AEP	0.357	0.074	20.73
AA	1.517	0.134	8.83
RTN	0.573	0.078	13.61
CPB	0.237	0.082	34.60
F	1.925	0.224	11.64
DIS	0.975	0.064	6.56
HPQ	1.015	0.119	11.72
BAX	0.422	0.087	20.62
XRX	1.314	0.106	8.07
WMB	1.069	0.137	12.82
WFC	1.103	0.119	10.79
WY	1.200	0.108	9.00
CSC	0.920	0.150	16.30
AVP	1.319	0.143	10.84
OMX	2.460	0.314	12.76
ATI	1.700	0.192	11.29
MCD	0.331	0.065	19.64
TYC	0.943	0.078	8.27
LB	1.000	0.185	18.50
JPM	1.093	0.107	9.79
TGT	0.732	0.099	13.52
WMT	0.215	0.078	36.28
AXP	1.577	0.125	7.93
INTC	0.829	0.092	11.10
BAC	1.775	0.184	10.37
MDT	0.732	0.084	11.48
FDX	1.049	0.091	8.67
CI	1.015	0.136	13.40
LTD	1.242	0.133	10.71
NSC	0.917	0.102	11.12
VZ	0.383	0.080	20.89
T	0.408	0.076	18.63
USB	0.816	0.094	11.52
HD	0.672	0.089	13.24
AIG	2.698	0.447	16.57
MS	1.255	0.145	11.55
C	2.012	0.177	8.80
BHI	1.167	0.142	12.17
CBS	1.889	0.091	4.82
CSCO	0.997	0.100	10.03
MEDI	5.516	1.966	35.64
AES	1.136	0.117	10.30
TWX	0.995	0.089	8.94
EP	0.976	0.171	17.52
ALL	1.152	0.092	7.99
MSS	-10.584	NaN	NaN
MLSS	0.787	1.225	155.65
HIG	2.263	0.189	8.35
LU	2.835	2.622	92.49
ROK	1.359	0.108	7.95
GS	1.163	0.114	9.80
UPS	0.680	0.075	11.03
CMCSA	0.859	0.091	10.59

Let us now consider sectoral variation in betas:

Sector-wise Summary of Estimated Betas
Sector	Average Beta	Average Std. Error	Average Std. Error %	n
Materials	1.639	0.131	7.99	5
Financials	1.537	0.163	10.31	11
Information Technology	1.247	0.348	20.44	13
Real Estate	1.200	0.108	9.00	1
Consumer Discretionary	1.136	0.287	26.67	11
Industrials	1.011	0.091	9.12	14
Energy	0.926	0.124	13.80	7
Communication Services	0.914	0.108	14.32	7
Consumer Staples	0.543	0.181	30.89	11
Utilities	0.476	0.087	23.91	5
Health Care	0.126	0.357	31.62	13

Our analysis reveals significant cross-sector variation in estimated beta coefficients, shedding light on the heterogeneous nature of systematic risk across industries. Sectors such as Materials (β = 1.639), Financials (β = 1.537), and Information Technology (β = 1.247) exhibit the highest average beta estimates. These elevated betas suggest that firms in these sectors tend to have above-average sensitivity to market-wide shocks, amplifying both upside potential and downside risk during economic fluctuations.

The cyclical nature of the Materials and Financials sectors makes them particularly vulnerable to changes in economic output, monetary policy, and commodity prices. For instance, demand for raw materials and lending activity both rise in expansions and fall during contractions, translating into higher co-movement with market returns. Similarly, Technology firms often derive value from future cash flows and are thus more sensitive to shifts in discount rates and investor sentiment—contributing to their higher beta values.

On the other end of the spectrum, sectors like Health Care (β = 0.126), Utilities (β = 0.476), and Consumer Staples (β = 0.543) exhibit significantly lower average betas. These sectors are traditionally seen as “defensive” because they offer goods and services that are considered essential, regardless of macroeconomic conditions. This results in more stable cash flows and less pronounced reactions to market-wide movements, making them relatively safe havens during periods of volatility or downturns.

An interesting nuance emerges when considering the standard error as a percentage of beta estimates. While the average betas of defensive sectors are lower, their standard errors (as a % of beta) are markedly higher — e.g., 31.62% for Health Care and 30.89% for Consumer Staples. This suggests greater uncertainty in measuring their true market exposure, possibly reflecting firm-level idiosyncrasies, regulatory shifts, or structural differences in business models. In contrast, cyclical sectors like Industrials, Materials, and Real Estate exhibit not only higher betas but also more statistically precise estimates, indicating a more robust and consistent relationship with market returns.

7.2 Cross-Sectional Regressions

Now, for part 2 - we have now estimated the Beta’s for each company. After this, for the Fama-Macbeth method, we will regress the monthly excess returns of each company against the market’s excess returns (where the market here is our universe of 100 companies), while controlling for EPS Surprise and Beta.

***Regressions by Crisis Period***

	Monthly Excess Return
	Pre-Crisis	Crisis	Post-Crisis

Beta (CAPM)	0.0001	0.013^**	-0.0002
	(0.005)	(0.005)	(0.002)
EPS Surprise	0.008	0.001^***	0.002^**
	(0.008)	(0.0002)	(0.001)
Constant	0.012^**	-0.010^*	0.013^***
	(0.006)	(0.005)	(0.002)
N	327	2,088	4,316
R²	0.003	0.020	0.001
Adjusted R²	-0.003	0.019	0.001
Residual Std. Error	0.062 (df = 324)	0.127 (df = 2085)	0.066 (df = 4313)
F Statistic	0.490 (df = 2; 324)	21.314^*** (df = 2; 2085)	2.892^* (df = 2; 4313)

p < .1; p < .05; p < .01

The table above presents the results of monthly Fama-MacBeth-style cross-sectional regressions where firm-level excess returns are regressed on market beta (CAPM) and earnings surprise (EPS Surprise). The regressions are conducted separately for three distinct economic periods: Pre-Crisis (2003–2006), Crisis (2007–2009), and Post-Crisis (2010–2013). This segmentation allows us to examine how the relationship between risk, information, and stock returns evolves across macroeconomic regimes.

The most striking observation is that during the Crisis period, both explanatory variables become statistically significant. The coefficient on Beta (0.013) is significant at the 5% level, and the coefficient on EPS Surprise (0.001) is highly significant at the 1% level. The overall model fit also improves markedly during this period, with an R² of 0.028 and a highly significant F-statistic (21.314), indicating strong joint explanatory power. This suggests that during periods of heightened uncertainty, such as the 2007–2009 financial crisis, investors become more responsive to both systematic risk and firm-specific information. In essence, risk and earnings surprises are more heavily priced into returns when markets are in turmoil, reflecting a flight to fundamentals and a shift toward information sensitivity under stress.

In contrast, both the Pre-Crisis and Post-Crisis periods exhibit much weaker relationships. In the Pre-Crisis window, neither Beta nor EPS Surprise is statistically significant, and the model explains virtually none of the variation in returns (R² = 0.003). This aligns with a period of widespread optimism, abundant liquidity, and lower return dispersion, where stock prices were likely driven by macro trends or investor sentiment rather than firm-specific fundamentals. Similarly, in the Post-Crisis era, although the F-statistic is marginally significant at the 10% level, the explanatory variables again lack statistical significance, and the R² drops back down to 0.001. This may indicate that once market conditions stabilized, the strong pricing of risk and information observed during the crisis gradually faded.

These findings provide strong empirical support for one of the central hypotheses of this capstone project: the relationship between stock returns, risk exposure, and earnings information is conditional on the macroeconomic environment. While fundamental signals like beta and EPS surprise are always present, their relevance fluctuates with investor psychology, market volatility, and systemic stress. Importantly, this confirms that cross-sectional return predictability is not static — it intensifies in periods of dislocation and recedes during calm. In this way, the results illustrate a broader insight: risk and information matter most precisely when markets are least predictable.

***Sector-wise Regressions (Top 5)***

	Monthly Excess Return
	Communication Services	Consumer Discretionary	Consumer Staples	Energy	Financials

Beta (CAPM)		0.002	-0.013^**	0.002	0.023
		(0.009)	(0.006)	(0.072)	(0.026)
EPS Surprise	0.087	0.026^***	0.045^***	-0.022	0.001^***
	(0.070)	(0.004)	(0.015)	(0.042)	(0.0004)
Constant	0.009	0.014	0.014^***	0.004	-0.029
	(0.006)	(0.010)	(0.003)	(0.080)	(0.057)
N	121	568	1,047	363	242
R²	0.013	0.072	0.015	0.001	0.052
Adjusted R²	0.004	0.068	0.013	-0.005	0.044
Residual Std. Error	0.065 (df = 119)	0.114 (df = 565)	0.059 (df = 1044)	0.099 (df = 360)	0.222 (df = 239)
F Statistic	1.527 (df = 1; 119)	21.803^*** (df = 2; 565)	8.125^*** (df = 2; 1044)	0.132 (df = 2; 360)	6.503^*** (df = 2; 239)

p < .1; p < .05; p < .01

This table shows how the relationship between monthly excess returns, beta (CAPM), and earnings surprises (EPS Surprise) varies across five major sectors. Notably, EPS Surprise is a strong and significant predictor of returns in Consumer Discretionary and Consumer Staples, suggesting investors in these sectors react more to firm-specific earnings news than to market-wide risk. In Consumer Staples, beta is even negatively significant, reinforcing its defensive profile where higher market exposure may be penalized.

By contrast, Communication Services and Energy show weak explanatory power — neither beta nor earnings surprises significantly predict returns, indicating these sectors may be driven by other factors like regulation or commodities. Financials show a small but significant effect for EPS Surprise, highlighting how firm-specific information can still matter, albeit more subtly.

Overall, these results validate our sector-based expansion. Return drivers differ sharply across industries, and understanding these differences adds depth and nuance to our Fama-MacBeth framework.

***Sector × Crisis Period Regressions (Sample)***

	Monthly Excess Return
	Communication Services - Pre-Crisis	Communication Services - Crisis	Communication Services - Post-Crisis	Consumer Discretionary - Pre-Crisis

Beta (CAPM)				-0.017
				(0.031)
EPS Surprise	-0.477	0.166	0.016	-0.011
	(0.366)	(0.142)	(0.086)	(0.035)
Constant	0.050	-0.0001	0.016^*	0.038
	(0.025)	(0.013)	(0.008)	(0.036)
N	6	36	79	36
R²	0.298	0.039	0.0005	0.010
Adjusted R²	0.122	0.010	-0.013	-0.050
Residual Std. Error	0.024 (df = 4)	0.077 (df = 34)	0.061 (df = 77)	0.090 (df = 33)
F Statistic	1.697 (df = 1; 4)	1.363 (df = 1; 34)	0.036 (df = 1; 77)	0.172 (df = 2; 33)

p < .1; p < .05; p < .01

This table presents sector × crisis period regressions and reveals how the impact of EPS surprises and market beta evolves across time and sectors. Within Communication Services, we observe a dramatic shift. In the Pre-Crisis period, EPS Surprise has a large negative coefficient (–0.477), though statistically insignificant due to the small sample size (N = 6). During the Crisis, the relationship turns positive (0.166), suggesting that earnings surprises were viewed more favorably amid market stress. In the Post-Crisis period, the coefficient flattens to 0.016, showing limited influence of EPS surprises, but a statistically significant positive constant emerges—possibly indicating sector-wide momentum or recovery effects unrelated to firm-specific news.

The Consumer Discretionary – Pre-Crisis model, meanwhile, shows a weak and statistically insignificant role for both Beta and EPS Surprise. Despite a decent sample size (N = 36), the low R² and adjusted R² suggest that market-wide and earnings-based factors were not strong predictors of returns in that sector and period.

Overall, these patterns highlight how market dynamics and investor behavior vary both across sectors and economic regimes. EPS Surprise appears more potent during times of volatility or stress in some sectors, while its relevance diminishes during stable periods—offering clear justification for the capstone’s expansion into crisis-period and sectoral subsamples.

Taken together, these cross-sectional regression results reinforce a central insight of this project: the pricing of risk and firm-specific information is highly conditional on both sectoral context and macroeconomic regime. While earnings surprises and market beta occasionally show statistically significant relationships with excess returns, their influence fluctuates markedly across sectors and crisis periods. During turbulent times, such as the 2007–2009 financial crisis, markets appear more sensitive to both systematic and idiosyncratic signals, whereas in calmer periods, these factors often lose explanatory power. Sectoral differences further amplify these effects, with some industries—like Consumer Staples and Discretionary—showing stronger responses to earnings news, while others remain inert. These findings not only validate the decision to expand the Fama-MacBeth framework but also emphasize the importance of tailoring asset pricing models to evolving market conditions and sectoral dynamics.

7.3 λ Visualized

In the Fama-Macbeth method, after estimating the beta coefficients that capture the sensitivities of asset returns to various risk factors, it becomes essential to understand their significance and impact on asset returns across different periods. This part of the analysis visualizes the lambda coefficients—representing the risk premiums associated with each factor.

The distribution of λ₂, which captures the cross-sectional price of market beta risk, is tightly clustered around zero, signaling that beta offers minimal explanatory power for returns across most months. This visual reinforces a central theme of this study: systematic risk is not consistently priced under normal conditions. The faint right skew — with a handful of elevated λ₂ values — hints at select stress periods where market beta briefly regains relevance, aligning with classical CAPM intuition that risk should be compensated in turbulent regimes. These occasional surges likely coincide with crisis windows (e.g., 2008–09), where investor demand for risk premiums spikes and systematic exposure becomes temporarily rewarded. Outside these episodes, however, beta remains largely ignored — a “phantom factor” in quiet markets.

Skewness might suggest that in some situations, higher beta (or higher market risk is associated with slightly higher returns, aligning with some risk-return tradeoff theories in finance like the Capital Asset Pricing Model (CAPM). A few outliers are seen on both the positive and negative sides.

The distribution of λ₃ (earnings surprise risk premiums) also clusters around zero, but with more spread than β estimates. This implies that while earnings surprises often have mild effects on excess returns, there are certain months — often coinciding with periods of economic dislocation — when these effects become strongly positive. The distribution’s leptokurtic shape reflects these episodic spikes in investor attention to firm-specific information.

The distribution of λ₃ coefficients — representing the cross-sectional risk premium on EPS surprises — is sharply peaked around zero, underscoring the generally modest and inconsistent influence of earnings surprises on excess returns once systematic risk is controlled for. Despite this central tendency, the distribution spans a meaningful range (approximately –0.5 to 0.5), signaling substantial temporal heterogeneity in the pricing of firm-specific information.

The elevated kurtosis reflects a regime in which earnings surprises are frequently ignored or weakly priced, consistent with periods of investor complacency or low information salience. However, the visible tails and occasional extreme values highlight episodes where earnings information becomes sharply priced, likely coinciding with market stress, heightened uncertainty, or shifts in macroeconomic expectations.

These dynamics lend strong support to the hypothesis that informational efficiency is state-dependent. In tranquil periods, markets may underreact to firm-level signals, whereas during crises or volatility spikes, investor attention reallocates toward fundamentals — causing surprises to command higher return premia. This episodic amplification of λ₃ reinforces the need for conditional asset pricing models that incorporate macro regime shifts and investor behavior under uncertainty.

7.4 λ T-Testing & Variance

T-test Results for Lambda2 (Beta Estimates
	Parameter	T-value	P-value	Mean	95% CI Lower	95% CI Upper
t	Beta Estimate	0.034108	0.9728477	0.0002105	-0.0120099	0.0124309

Across the full sample of sector- and crisis-aware cross-sectional regressions, the estimated λ₂ (risk premium on market beta) is not statistically distinguishable from zero. The t-test confirms that market beta has no systematic pricing power in this setting — echoing earlier findings of its limited predictive value.

This reinforces one of the core findings of the capstone: information asymmetry and investor underreaction to firm-specific news remain exploitable in asset pricing — particularly when viewed through a sectoral lens.

T-test Results for Lambda3 (EPS Surprise)
	Parameter	T-value	P-value	Mean Estimate	95% CI Lower	95% CI Upper
t	EPS Surprise	2.531488	0.0126506	0.0180438	0.0039314	0.0321562

The P-value of 0.01265 is below the typical significance threshold of 0.05, suggesting that there is statistical evidence to reject the null hypothesis that the mean effect of the EPS surprise on excess returns is zero.

A confidence interval of [0.003931354, 0.032156180] does not include zero and is entirely positive, reinforcing the significant positive effect of EPS surprise on excess returns.

The t-value of 2.5315 indicates that the sample mean is more than two standard deviations away from zero.

7.5 λ Standard Errors

Variance & Standard Error for λ2 and λ3
Parameter	Variance	Standard Error
λ2 (Beta Estimate)	3.78e-05	0.0061466
λ3 (EPS Surprise)	5.04e-05	0.0070982

For λ2, we see a very small variance value - there is low variability in impact of market beta on excess returns over the months.

Additionally, a smaller standard error suggests higher precision of the beta estimates across the sample. This reinforces the notion that the λ2 estimates are quite consistent, though as earlier t-test results indicated, their overall effect may not be statistically significant (given the p-value was high).

For λ3, we see slightly more variability - and a larger standard error reflects this.

We now calculate the average of the coefficients, and conduct t-tests to check significance, the results of which are printed below.

T-Test Results for λ Parameters
Term	t_Value	p_Value	Confidence_Interval
λ2 (Beta Estimate)	0.034108	0.9728477	-0.012 to 0.012
λ3 (EPS Surprise)	2.531488	0.0126506	0.004 to 0.032

As we can see, for the Beta Estimates, there is a very low t-value and an ample sample to estimate variability. Additionally, there is a very high p-value, suggesting that the market beta does not significantly affect excess returns.

For the EPS Surprise λ, the mean of the EPS surprise estimates is significantly above zero. The p-value is below the significance level of 0.05, providing evidence to reject the null hypothesis that the mean of EPS surprise estimates is zero. The average effect of EPS surprises is significantly positive; thus the confidence interval does not include zero and is positive.

7.6 Sectoral Conditional Alpha Map

To summarize the relationship between market beta and earnings surprises across sectors, the chart below maps how each sector prices these two core dimensions of return. It reveals that defensive sectors like Healthcare and Staples price firm-specific information more consistently, while cyclicals like Financials and Energy remain heavily driven by market-wide beta. This visual reinforces the study’s central insight: alpha is conditional—on both sector and regime.

This chart maps how sector returns are driven by market risk (λ₂) and earnings surprises (λ₃). Defensive sectors like Healthcare and Consumer Staples price information more than macro moves — ideal for alpha. In contrast, Financials and Energy are more market-driven, offering less alpha from surprises. This supports the thesis that alpha is conditional on both sector and macro regime.

8 Results

This study finds that the market’s reaction to earnings surprises is not uniform but conditioned by both sectoral identity and macroeconomic regime. Using Fama-French and Fama-MacBeth frameworks, I show that earnings surprises carry a significant return premium during crisis periods—especially in defensive sectors like Healthcare and Consumer Staples.

In contrast, market beta is only priced meaningfully during periods of high volatility, such as the 2007–2009 financial crisis. Long-short strategies based on surprise signals perform best in resilient sectors during downturns, but lose explanatory power in stable conditions.

These results underscore a key insight: investors process firm-specific news differently depending on both the economic environment and sectoral context. Portfolio strategies must therefore be crisis-aware and sector-specific to consistently extract alpha.

8.1 🔍 Summary of Findings

Finding	Interpretation	Significance
EPS Surprises are significantly priced during crises	Earnings news is amplified in volatile environments	Investors crave clarity in uncertainty
Market beta only matters in crises	CAPM-style risk premiums are dormant in calm markets	Rethink risk models outside of recessions
Defensive sectors price surprises more consistently	Earnings signals in Staples and Healthcare are more predictive	Stronger investor trust and less cyclicality
Long-short strategies work best in downturns	Surprise-based alpha is conditional	Timing and sector targeting matter
Weak predictive power in stable periods	Signals get lost in macro calm or euphoria	Markets price narratives more than fundamentals

8.2 💡 Real-World Implications

8.2.1 📊 Portfolio Strategy Design

Use-case: Hedge funds deploying post-earnings drift (PEAD) strategies.
Insight: Focus on defensive sectors during recessions — this is where earnings surprises are most predictive.
Example: In 2025, long positions in Healthcare and Consumer Staples yield better alpha than Tech.

8.2.2 🛡️ Risk Management

Use-case: Asset managers evaluating exposure to high-beta names.
Insight: Beta only matters in crises. Outside of downturns, it’s not a priced risk.
Example: Factor timing matters. High-beta exposure pays off in expansionary phases, but quality, size, and defensiveness take over when the cycle turns.

8.2.3 🔄 Sector Rotation Models

Use-case: Macro funds reallocating based on policy changes and earnings cycles.
Insight: Sectoral reactions vary by crisis type — cyclicals may underreact or misprice earnings.
Example: In 2008, Financials imploded on earnings misses. In 2025, Energy is thriving due to geopolitical supply shocks.

8.2.4 📉 Sell-Side Equity Research

Use-case: Equity analysts forecasting post-earnings drift.
Insight: In volatile macro conditions, positive earnings surprises in defensive sectors are more likely to yield sustained price reactions.
Example: An earnings beat from PepsiCo in Q1 2025 may matter more than one from Ford.

8.2.5 🧮 Policy Advisory

Use-case: Central banks monitoring asset price fragility.
Insight: Strong pricing of EPS surprises = heightened market sensitivity = early warning of instability.
Example: Central banks can use this as a signal of stress to preempt contagion or illiquidity events.

8.3 📌 Takeaway: Conditional Alpha and Crisis-Aware Investing

The evidence affirms a core truth: earnings surprises are conditionally priced, not uniformly. Their impact intensifies during crises and in specific sectors. Meanwhile, market beta is largely irrelevant unless fear dominates the market.

As investors rotate toward defensives in 2025, this study offers a practical roadmap for navigating uncertainty:

Target resilient sectors when volatility spikes.
Use earnings surprises tactically, not uniformly.
Ditch static models — adopt macro- and sector-aware frameworks.

💬 In modern markets, alpha isn’t dead — it’s just hiding where fear lives and where narrative doesn’t.

9 Study Limitations

While this study provides a robust empirical examination of how sectoral characteristics and macroeconomic regimes shape investor reactions to earnings surprises, several limitations must be acknowledged. First, the dataset spans 2003 to 2013, a period defined by the global financial crisis and its aftermath. Although this timeline is ideal for stress-testing hypotheses, it may not fully reflect more recent structural shifts in market behavior, such as the rise of algorithmic trading, real-time sentiment flows, and post-pandemic macro dynamics. Second, the classification of firms using GICS sector codes, while analytically convenient, imposes a static structure on a fluid economy. Many modern firms operate across sectors or have business models that defy traditional categorization, which can dilute the explanatory power of sector-based analysis.

Additionally, the study employs a narrow definition of earnings surprises based solely on deviations from consensus analyst forecasts. This approach overlooks the qualitative elements that increasingly shape investor response—management tone, forward guidance, or geopolitical context. Moreover, the use of beta and standard asset pricing factors, though well-established, may not fully capture the breadth of risks priced in today’s markets, particularly during periods of heightened uncertainty when narratives, liquidity, and behavioral biases dominate. Finally, survivorship bias remains a concern, as the dataset primarily reflects firms that withstood crisis periods. The exclusion of firms that failed or were acquired may understate the true variability of sectoral responses to shocks.

Despite these constraints, the study offers meaningful insights into the conditional nature of earnings surprise pricing. Future research may benefit from incorporating alternative surprise metrics, dynamic sector definitions, and microstructure-level data to capture a more complete picture of how investors interpret firm-level information under different macro regimes.

10 Strategic Outlook & Conclusion

10.1 Macroeconomic and Geopolitical Crosscurrents

The re-election of Donald Trump has ushered in aggressive protectionist trade policies that blindsided investors. Swift tariff escalations against not just China but close U.S. partners (Canada, Mexico, Europe) have “spooked investors” and sparked a sharp equity sell-off. In mid-March, the S&P 500 plunged into a correction, erasing over $4 trillion in market value from its February peak. What began as optimism over tax cuts and deregulation quickly morphed into fear as tariff volleys and unpredictable policy shifts drove a “big sentiment shift” on Wall Street. Business confidence has deteriorated under record-high trade policy uncertainty with many firms delaying capital investments and hiring plans until the outlook stabilizes. This protectionist pivot amounts to the largest peacetime tariff shock in modern U.S. history, effectively a tax increase that could eat into corporate profits and consumer spending. In short, policy uncertainty is imposing a risk premium on markets that cannot be ignored.

Currency markets have been whipsawed by 2025’s geopolitical tensions. Contrary to many expectations, the U.S. dollar weakened against almost all major currencies in Q1 (with the sole exception of Canada’s). Investors normally view U.S. tariffs as dollar-supportive, but when aimed at close allies the effect backfired – undermining confidence in U.S. policy and growth. Recession worries and trade rifts have instead driven money into other currencies: the euro is up ~5% this quarter, hitting its highest level since the U.S. election. Japan’s yen has also surged (~6% stronger vs USD) on safe-haven inflows and a more hawkish Bank of Japan. In fact, speculators have now built record-long positions expecting further yen gains, anticipating that rising wages and inflation in Japan will push the BOJ to keep raising rates. Meanwhile, emerging-market FX has been turbulent. China’s yuan, widely expected to weaken under new U.S. tariffs, has strengthened to around 7.25 per dollar as Beijing maintains stability and other Asian currencies rise even more. The Mexican peso rebounded ~5% from its panic lows after some tariff suspensions, but volatility remains elevated. Overall, currency markets have become the financial system’s “Achilles’ heel” – with KKR warning that trade disputes and fiscal imbalances could trigger “potential volatility shocks” in FX going forward. For global investors, this means larger currency swings (and potential losses) if portfolios aren’t hedged for USD weakness or EMFX instability.

One of the starkest anomalies of early 2025 has been the weakness of traditional safe-haven assets in the face of uncertainty. U.S. Treasuries – normally a port in the storm – have traded more like risky assets of late. Initially, the tariff-driven equity rout saw investors pile into Treasuries (driving yields down), but this “flight to quality” abruptly reversed in April. In a flashback to March 2020’s dash-for-cash, a “wave of selling” hit even government bonds, sending 10-year yields soaring 17 basis points in a single day despite stocks still falling. Such violent Treasury swings (intraday yield ranges of ~35 bps, the widest in decades) signal strained liquidity and possible forced selling by funds meeting margin calls

Several factors are undermining the usual safe-haven bid. First, persistent inflation and stagflation fears are keeping bond investors skittish – short-term inflation expectations have risen, and U.S. breakeven rates moved higher in Q1. This has pushed up real yields and the term premium on U.S. debt to the highest in about ten years. In other words, investors are demanding greater compensation (yield) to hold Treasuries given inflation and policy uncertainty. Second, fiscal concerns are mounting. Moody’s recently warned that unfunded tax cuts and sustained high tariffs could worsen U.S. debt dynamics, potentially leading to “higher interest rates and worsening debt” that erode America’s credit standing. (The U.S. already carries a negative outlook on its AAA rating, and Fitch’s 2023 downgrade looms in the background.) These fiscal and policy risks have dented confidence in U.S. government bonds as the ultimate safe asset. Instead, some investors have rotated into alternative havens: gold, for example, has rocketed to over $3,000/oz – a fresh record high, reflecting demand for inflation-proof, non-dollar stores of value. Likewise, traditional “safe” currencies like the Swiss franc and Japanese yen have attracted inflows.

💬 The key takeaway is that high inflation and deficit worries are raising risk premia on sovereign debt rather than lowering them. When both stocks and bonds sell off together (a breakdown of the usual negative correlation), balanced portfolios face a much tougher environment.

As inflation stayed sticky and tariffs hit confidence, investors aggressively repositioned into areas seen as more resilient. High-valuation growth and Big Tech names (the “Magnificent Seven”) that led 2023’s rally have stumbled; this elite cohort fell over 16% as a group in Q1, suffering a “rude awakening” as lofty AI-driven expectations met a harsher reality. In their place, value and defensive stocks have taken the lead. Traditional safe sectors like Utilities, Consumer Staples, Healthcare, and dividend-paying aristocrats held up relatively well or gained, while cyclicals and rate-sensitive growth lagged. Indeed, defensive sectors outperformed the broader market throughout Q1 “as investors sought out safe havens.” Utilities and real estate stocks actually rose in Europe during the post-tariff selloff, and defense/aerospace shares surged on the back of higher military spending.

U.S. financials sagged under recession fears and yield-curve pressures. This pattern is reminiscent of past crises: investors rotate out of high-beta, growth-oriented segments and into quality, income, and stability when uncertainty strikes. Notably, energy stocks have been a wild card – often considered cyclical, oil & gas shares actually climbed in Q1 (with Energy +9% in the S&P) as oil prices were buoyed by supply concerns and geopolitical risk. The U.S. market’s leadership has clearly narrowed and shifted away from the tech-centric dominance of recent years. In fact, U.S. equities underperformed foreign markets in Q1 by the widest margin in 15+ years.

All of this underscores how quickly investor sentiment can flip in today’s environment – from exuberance to risk-aversion – and how macro policy surprises are dictating the rotations. With inflation not yet vanquished and policy in flux, we should expect further gyrations in market leadership as the year progresses.

Focus on companies with strong balance sheets, durable cash flows, and pricing power.
Defensive sectors like Consumer Staples, Healthcare, Utilities, and Defense outperformed in 2025. Quality/value stocks—such as Dividend Aristocrats and free cash flow leaders—offer resilience if stagflation or a hard landing unfolds.
International diversification has paid off.
While U.S. equities fell in Q1, European and emerging markets (e.g., Brazil, China) gained due to lower valuations and local catalysts. With the dollar weakening and FX volatility high, hedging currency risk or tilting toward safe-haven currencies (JPY, CHF) is prudent.
Currency swings are a major risk and opportunity.
The yen’s rise provides a hedge against U.S. slowdown, while unhedged currency exposure can either amplify or erode returns. Use FX futures, ETFs, or local-currency share classes to avoid excessive single-currency exposure.
The 60/40 portfolio model is under pressure.

Stocks and bonds have sometimes declined together, weakening traditional diversification. Alternatives like gold (now above $3,000/oz), TIPS, commodity equities, floating-rate debt, and short-term Treasuries offer better downside protection. Policymakers should monitor Treasury market liquidity and be prepared for intervention if needed.

Policy credibility is critical.
Central banks must clarify how they’ll respond to weakening growth amid persistent inflation. Blanket tariffs may backfire—reinvesting tariff revenue into infrastructure, R&D, and workforce development is a smarter long-term play. Coordinated fiscal and monetary action is key to restoring market confidence.

10.2 Lessons from “Earnings Surprises, Sector Dynamics, and Crisis Effects”

Today’s volatile macro environment adds real-time relevance to my capstone, “Earnings Surprises, Sector Dynamics, and Crisis Effects.” We’re seeing firsthand how macro turmoil is distorting the pricing of firm-level shocks.

In calmer markets, earnings beats drive strong and sustained price moves. In 2025, however, this dynamic is subdued. Even positive surprises often yield fleeting or muted stock reactions, as investors prioritize macro headlines—Fed moves, trade wars—over quarterly fundamentals. One-day returns are increasingly decoupled from earnings results, with forward guidance and geopolitical commentary driving sentiment.

The crisis playbook still holds: defensives (Staples, Utilities, Healthcare) have outperformed, offering stability amid uncertainty. Cyclicals—autos, luxury goods, industrials—have lagged, particularly after tariff shocks. However, this crisis presents twists: energy stocks have outperformed due to supply shocks, not demand, and tech is behaving inconsistently, with some mega-caps acting defensively while others face steep sell-offs. The lesson: sector effects depend on the type of crisis.

Earnings-based long-short strategies, which usually profit from post-surprise drift, are under strain. In crisis periods, alpha often vanishes as macro factors swamp idiosyncratic news. Stocks that beat earnings may still fall if the sector narrative is bearish. To adapt, managers must tighten risk controls, shorten holding periods, and consider macro hedges. Sector context matters more than ever—defensive surprise winners are more likely to outperform than cyclical ones.

Even if short-term reactions are noisy, earnings calls reveal forward-looking insights. In Q1 2025, commentary has focused on tariffs, supply chains, and consumer weakness—leading to broad downward revisions. While the market may ignore beats or misses in the moment, aggregate earnings trends still inform policy moves and recession risk. For investors and policymakers alike, earnings remain a key window into the economy’s resilience.

In summary, the first quarter of 2025 has delivered a masterclass in how macroeconomic and geopolitical factors can upend conventional market wisdom. Protectionist policies, currency volatility, and inflation have created a challenging mosaic for investors: one where stocks and bonds can fall together, where yesterday’s winners (Big Tech) become today’s losers, and where defensive positioning and agility trump complacent buy-and-hold strategies. Yet, within this turbulence lie opportunities. The prudent investor or policy advisor will heed these signals – rotating portfolios to sturdier ground, hedging the new risks, and staying adaptive to fast-changing information. The current environment, as tumultuous as it is, validates many of the patterns examined in “Earnings Surprises, Sector Dynamics, and Crisis Effects.” We see that crisis-driven markets reward safety and punish vulnerability, that earnings news is filtered through a macro lens, and that strategies must evolve when the old correlations break down. By integrating these lessons, equity investors and portfolio managers can navigate the rest of 2025 with a clearer strategic vision, while policymakers can glean how their decisions ripple through sectors and sentiment. The road ahead will likely feature continued volatility, but with volatility comes opportunity for those prepared to respond with discipline and foresight. In essence, surviving and thriving in 2025 will require staying grounded in fundamentals but nimble in execution – a combination of steady hands on the wheel and eyes wide open to the cross-currents swirling in this new market regime.

10.3 Conclusion

This capstone set out to answer a fundamental question: how do investors price earnings surprises across different sectors and macroeconomic conditions? The evidence is clear—there is no uniform answer. Instead, the pricing of earnings information is conditional, sector-sensitive, and crisis-contingent.

Through a blend of CAPM, Fama-French, and Fama-MacBeth models, this study uncovered that beta is not consistently priced, and its explanatory power evaporates outside of crisis periods. In contrast, earnings surprises carry a significant and persistent return premium—especially in volatile regimes and resilient sectors. Long-short strategies built on these signals perform best in defensive industries during stress, validating the thesis that macro context and sectoral positioning jointly shape investor reactions.

This insight resonates powerfully in today’s fractured 2025 markets. Just as during the 2007–2009 crisis, investors now favor quality, durability, and predictability. Amid trade wars, stagflation fears, and synchronized stock-bond drawdowns, the same sectors that showed resilience in the past—healthcare, staples, and utilities—are outperforming once again. Meanwhile, tech and cyclicals show more fragile, uneven responses to even strong earnings beats. Importantly, this project also highlights how alpha evaporates when the market narrative overwhelms firm-specific news. In 2025, just as in 2008, investors discount good earnings if the macro story is grim. Strategies that ignore this conditionality are likely to underperform.

In sum, this capstone affirms a core truth: fundamentals matter most when uncertainty is highest. Crisis periods amplify signal-to-noise ratios, rewarding investors who combine rigorous factor analysis with contextual judgment. The path forward—whether for policymakers or asset managers—requires a dual lens: one that sees the signal in earnings surprises, but also adjusts the filter based on sectoral dynamics and economic stress.

This is not just a study of past data. It’s a roadmap for navigating market regimes that reward precision, punish complacency, and demand nuance. In an era where old correlations are breaking down, this research offers a framework to rebuild them—smarter, sharper, and more resilient.

11 Works Cited

News & Market Commentary (2025)

Reuters. (2025, April 30). Annual gold price forecast tops $3,000 for first time: Reuters poll. Retrieved from https://www.reuters.com/markets/commodities/annual-gold-price-forecast-tops-3000-first-time-2025-04-30/

MarketWatch. (2025, April 30). Timeline: How Trump’s first 100 days shook stocks, bonds and other financial markets. Retrieved from https://www.marketwatch.com/story/timeline-how-trumps-first-100-days-shook-stocks-bonds-and-other-financial-markets-93e75b6d

New York Post. (2025, April 30). US economy unexpectedly shrank in first three months of 2025 on import rush ahead of Trump tariffs. Retrieved from https://nypost.com/2025/04/30/business/us-economy-shrank-in-first-three-months-of-2025-as-trump-tariffs-hit-businesses/

Business Insider. (2025, April 30). Gold will keep setting records with a recession more likely than people think, Goldman says. Retrieved from https://www.businessinsider.com/gold-price-records-recession-central-banks-us-government-risks-economy-2025-4

Associated Press. (2025, April 30). Tariff turmoil prompts cloudy forecasts from companies for the year ahead. Retrieved from https://apnews.com/article/93474e22fcf883d9040cd8ac1adbfc67

Reuters. (2025, March 10). US stock market loses $4 trillion in value as Trump plows ahead on tariffs. Retrieved from https://www.reuters.com/markets/us/investors-flee-equities-trump-driven-uncertainty-sparks-economic-worry-2025-03-10/

Reuters. (2025, March 14). The big currency winners of 2025 so far do not include the dollar. Retrieved from https://www.reuters.com/markets/currencies/big-currency-winners-2025-so-far-do-not-include-dollar-2025-03-14/

Western Asset. (2025, April 7). US Treasuries—Safe-Haven Status Reaffirmed, with a Caveat. Retrieved from https://www.westernasset.com/us/en/research/blog/us-treasuries-safe-haven-status-reaffirmed-with-a-caveat-2025-04-07.cfm

Euronews. (2025, April 9). Why are US Treasury bonds losing their safe-haven status in dramatic sell-off. Retrieved from https://www.euronews.com/business/2025/04/09/why-are-us-treasury-bonds-losing-their-safe-haven-status-in-dramatic-sell-off

Investor’s Business Daily. (2025, April 29). Magnificent Seven Stocks: Amazon, Nvidia, Tesla Fall. Retrieved from https://www.investors.com/research/magnificent-seven-stocks-april-2025/

Investopedia. (2025, March 31). Why the Magnificent Seven Stocks Just Had Their Worst Month and Quarter on Record. Retrieved from https://www.investopedia.com/magnificent-seven-stocks-worst-month-quarter-on-record-q1-2025-11706435

CBS News. (2025, March 14). Gold’s price breaks record $3,000 per ounce: Everything to know now. Retrieved from https://www.cbsnews.com/news/golds-price-breaks-record-3000-per-ounce-everything-to-know-now-march-2025/

The Guardian. (2025, March 18). Gold hits record high over $3000 amid rising geopolitical tensions. Retrieved from https://www.theguardian.com/business/live/2025/mar/18/gold-hits-record-high-over-3000-geopolitical-tensions-weakening-us-dollar-oil-stock-markets-inflation-business-live-news

Wikipedia. (2025, April 30). Second presidency of Donald Trump. Retrieved from https://en.wikipedia.org/wiki/Second_presidency_of_Donald_Trump

Academic & Peer-Reviewed Sources

Ball, Ray, and Philip Brown. “An Empirical Evaluation of Accounting Income Numbers.” Journal of Accounting Research, vol. 6, no. 2, 1968, pp. 159–178. JSTOR, doi:10.2307/2490232.

Bernard, Victor L., and Jacob K. Thomas. “Post-Earnings-Announcement Drift: Delayed Price Response or Risk Premium?” Journal of Accounting Research, vol. 27, 1990, pp. 1–36. JSTOR, doi:10.2307/2491062.

Barberis, Nicholas, Andrei Shleifer, and Robert Vishny. “A Model of Investor Sentiment.” Journal of Financial Economics, vol. 49, no. 3, 1998, pp. 307–343. Elsevier, doi:10.1016/S0304-405X(98)00027-0.

Doyle, Jeffrey T., Richard J. Lundholm, and Mark T. Soliman. “The Extreme Future Stock Returns Following I/B/E/S Earnings Surprises.” Journal of Accounting Research, vol. 44, no. 5, 2006, pp. 849–887. JSTOR, doi:10.1111/j.1475-679X.2006.00220.x.

Skinner, Douglas J., and Richard G. Sloan. “Earnings Surprises, Growth Expectations, and Stock Returns or Don’t Let an Earnings Torpedo Sink Your Portfolio.” Review of Accounting Studies, vol. 7, no. 2–3, 2002, pp. 289–312. Springer, doi:10.1023/A:1020294523516.

Brown, Lawrence D., Andrew C. Call, Michael B. Clement, and Nathan Y. Sharp. “Inside the ‘Black Box’ of Sell-Side Financial Analysts.” Journal of Accounting Research, vol. 46, no. 4, 2008, pp. 667–708. Wiley, doi:10.1111/j.1475-679X.2008.00293.x.

Keung, Ethel, Steven C. Lin, and Kim Wai Wu. “Earnings Surprises, Investor Sentiment, and Stock Returns.” Journal of Banking & Finance, vol. 34, no. 10, 2010, pp. 2352–2362. Elsevier, doi:10.1016/j.jbankfin.2010.02.008.

Fama, Eugene F., and Kenneth R. French. “Common Risk Factors in the Returns on Stocks and Bonds.” Journal of Financial Economics, vol. 33, no. 1, 1993, pp. 3–56. Elsevier, doi:10.1016/0304-405X(93)90023-5.

Brunnermeier, Markus K. “Deciphering the Liquidity and Credit Crunch 2007–2008.” Journal of Economic Perspectives, vol. 23, no. 1, 2009, pp. 77–100. American Economic Association, doi:10.1257/jep.23.1.77.

Gorton, Gary B. “The Panic of 2007.” National Bureau of Economic Research Working Paper Series, no. 14358, 2008, pp. 1–50. doi:10.3386/w14358.

Bharath, Sreedhar T., and Tyler Shumway. “Forecasting Default with the Merton Distance to Default Model.” Review of Financial Studies, vol. 21, no. 3, 2008, pp. 1339–1369. Oxford University Press, doi:10.1093/rfs/hhn044.

Fahlenbrach, Rüdiger, and René M. Stulz. “Bank CEO Incentives and the Credit Crisis.” Journal of Financial Economics, vol. 99, no. 1, 2011, pp. 11–26. Elsevier, doi:10.1016/j.jfineco.2010.08.010.

Dichev, Ilia D., and Joseph D. Piotroski. “The Long-Run Stock Returns Following Bond Ratings Changes.” Journal of Finance, vol. 56, no. 1, 2001, pp. 173–203. Wiley, doi:10.1111/0022-1082.00322.

Campbell, John Y., Stefano Giglio, and Christopher Polk. “Hard Times.” Review of Asset Pricing Studies, vol. 2, no. 1, 2008, pp. 95–132. Oxford University Press, doi:10.1093/rapstu/rst004.

Chen, Long, and Lu Zhang. “Risk, Growth, and Cross-Sectional Predictability of Earnings and Returns.” Journal of Financial Economics, vol. 129, no. 2, 2018, pp. 498–519. Elsevier, doi:10.1016/j.jfineco.2018.05.006.

Abarbanell, Jeffrey S., and David Park. “Earnings Surprises and Information-Processing Biases: The Impact of Investor Sophistication.” Contemporary Accounting Research, vol. 33, no. 3, 2016, pp. 1014–1038. Wiley, doi:10.1111/1911-3846.12217.

Flight to Fundamentals: Earnings Surprise Pricing When the Music Stops

Sectoral and Crisis-Period Conditioning in the Post-Earnings Drift

Abdullah Ahmed - Capstone

May 2025