Motivations

Who really moves the market?

Financial markets are often portrayed as being ruled by a handful of giants.
But when an entire sector shifts, is it truly the leaders pulling the strings, or is that impression simply a byproduct of size?

In this project, we treat the Nasdaq not as a single voice, but as a set of sectors where firms have very different weights and roles. Rather than assuming large caps dominate, we test it: we rank companies within each sector and track how returns and volatility propagate over time.

Think of each sector as a group moving together — but not everyone steps at the same time. Below, we explore who tends to move first and who follows. We also compare two ways of “building the market”: one where big companies weigh more, and one where every company counts equally. Side by side, this tells us whether performance comes from a few giants or from the whole pack.

Do signals originate from industry leaders and cascade downward? Or do they form collectively, across companies of all sizes?

By following the data, and letting the charts speak, this project offers a clearer view of market power, sector dynamics, and how information actually diffuses through the Nasdaq.

Data Overview

• NASDAQ Stock Market Dataset (Kaggle): daily OHLCV price data for stocks and ETFs + unified metadata (symbols_valid_meta.csv)

• External Data: NASDAQ Company List (Kaggle): companies only, with Market Cap, Sector/Industry, and IPO Year (companylist.csv)

External Dataset overview:

Click a sector bubble to reveal the full list of companies in that sector.

---

Exploring The Market

Let’s dive into our data ! We start by stripping the dataset down to what matters for our question: single firms.
For the first part of the project, we therefore exclude ETFs (we only use some of them in the last section) and focus on individual NASDAQ-listed companies.

To enrich the metadata dataset, we merge it with the external company list, adding key descriptors for the 2,074 remaining stocks (market cap, IPO year, sector). With the universe cleaned and mapped, we can finally “see” what the NASDAQ looks like once split into sectors:

A first striking pattern is how uneven size is within sectors: the top 5% firms by market cap often capture a large share of the sector’s total value. The magnitude differs across sectors, but the takeaway is consistent: sectors may contain many firms, yet value can still be concentrated in a handful of giants (in terms of market cap), exactly the kind of structure where “leadership” effects could emerge.

To complement this snapshot of “size”, we use IPO year as a proxy for sector growth through new entrants: the stacked area chart tracks the cumulative number of listed companies by sector. This helps compare sector dynamics as some expand steadily while others remain relatively stable. This also motivates that there must be additional ranking dimensions beyond market cap (e.g., how established a company is, and how “crowded” its sector is).

This section sets the stage for the next steps, as we now have handful of elements to hierarchize the companies within each sector, to then be able to process our leadership analysis, testing whether influence follows the same concentration patterns (or not).

---

Hierarchizing Market Participants

How should companies be sectorized and hierarchized?

–> Building the Power Score: Beyond Market Cap

To answer whether the most powerful firms are also the most influential, we need a way to rank companies that goes beyond simple market capitalization. Size matters, but influence is multidimensional.

---

The Five Pillars of Power

We construct a Power Score that integrates five key financial indicators. Each metric captures a different dimension of market strength:

---

How We Weight Each Factor

Not all metrics are created equal. Based on financial research and market dynamics, we assign weights that reflect each factor’s importance in determining market power.

The Power Score is constructed as a weighted combination of five standardized indicators:

Metric	Weight	Interpretation
Market Capitalization	0.40	Size and financial solidity
Trading Volume	0.25	Liquidity and visibility
Mean Return	0.15	Performance over time
Inverse Volatility	0.10	Price stability
Age Since IPO	0.10	Firm maturity and credibility

Weights sum to one and ensure that no single dimension dominates the ranking.

---

Visualizing the Power Hierarchy

Explore the top companies by Power Score across different sectors. Use the filters to dive deeper:

---

Real-World Context

---

What’s Next?

This Power Score becomes the foundation for our leadership analysis. Once we’ve ranked companies within each sector, we can investigate:

• Do high Power Score companies lead price movements?
• Is influence proportional to power, or are there surprises?
• How consistent are leadership patterns over time?

---

Technical Details: Standardization & Weight Selection

Z-Score Normalization

All quantitative variables are standardized to ensure comparability:

\[ x_i' = \frac{x_i - \mu_x}{\sigma_x} \]

where \(\mu_x , \sigma_x\) are the mean and standard deviation across all companies.

Weight Constraints

The weights satisfy:

\[ w_k \ge 0, \quad \sum_{k=1}^{5} w_k = 1 \]

Complete Formula

The full Power Score formula:

\[ S_i = w_1 \cdot Cap_i + w_2 \cdot Vol_i + w_3 \cdot Ret_i + w_4 \cdot (1 - Vol_i) + w_5 \cdot Age_i \] with \[ w_1 = 0.40, w_2 = 0.25, w_3 = 0.15, w_4 = 0.10, w_5 = 0.10. \]

---

Analysis of Stock Leadership

---

Rather than assuming leadership a priori, we let the data reveal which companies tend to move first.

We conclude by revisiting our main research questions and summarizing the empirical insights provided by the leader–follower analysis.

---

How do we define a “leader” and a “follower” in stock movements?

A leader is a stock whose past daily returns improve the prediction of another stock’s future returns. A follower is a stock that reacts with a short delay to movements in the leader.

In practice, leadership is assigned only when two conditions are met:

• a statistically significant lead–lag correlation at a positive time shift, and
• a Granger causality test confirming predictive precedence in one direction.

---

How can directional influence be detected within sectors?

Directional influence is detected through a two-step statistical pipeline. First, cross-correlation analysis identifies candidate lead–lag relationships. Second, Granger causality tests verify whether past returns of one stock improve prediction of another beyond its own history.

Only relationships supported by both steps are retained, ensuring that detected links reflect directional predictability rather than simple co-movement.

---

Can daily return time series reveal short-term influence?

Yes. Daily return series are sufficiently granular to reveal lead–lag effects over horizons of a few trading days. While these effects are moderate in magnitude, they are statistically meaningful and consistently detected across multiple sectors.

---

Are leader–follower dynamics consistent across sectors?

No. Leadership patterns are strongly sector-dependent. Some sectors exhibit structured hierarchies with multiple leaders, others display only a few isolated leader–follower pairs, and several sectors show minimal detectable leadership.

Sectors such as Health Care, Energy, and Transportation tend to form denser and more structured leadership networks, suggesting faster or more coordinated information transmission within these industries.

---

Mathematical details

Lead–Lag Cross-Correlation

For each ordered pair of stocks \( (i,j) \) within a sector, we test whether movements in \( i \) tend to precede movements in \( j \) by computing:

\[ \rho_{ij}(k) = \mathrm{Corr}\!\big(r_{i,t},\, r_{j,t+k}\big), \qquad k = 1,\dots,7 \]

If the strongest correlation occurs at a positive lag \( k>0 \), stock \( i \) is treated as a candidate leader of stock \( j \).

Granger Causality Test

To establish directionality, we compare a baseline autoregressive model to an augmented model that includes lagged returns of the leader:

\[ r_{j,t} = \alpha + \sum_{\ell=1}^{p} \beta_\ell r_{j,t-\ell} + \sum_{\ell=1}^{p} \gamma_\ell r_{i,t-\ell} + \varepsilon_t, \qquad p = 3 \]

A directed link \( i \rightarrow j \) is retained only if the coefficients \( \gamma_\ell \) are jointly significant, and no reverse causality is detected.

Interactive Sector Leadership Heatmap

This heatmap shows, for each sector, how strongly each stock (rows) appears to lead others (columns) based on our cross-correlation.

Select sector:

Leader–Follower Network Graph (per sector)

Each node is a stock. Arrows point from leader → follower. Edge thickness reflects the strength of the relationship (|correlation|).

Select sector:

Sector-by-Sector Leadership Insights

The analysis identifies sparse but structured leader–follower relationships within sectors. Rather than dense interactions, leadership effects concentrate around a limited number of firms and propagate in clearly defined directions.

Basic Industries

Leadership is shared among several industrial firms, with Steel Dynamics (STLD), WD-40 Company (WDFC), and Matrix Service (MTRX) acting as upstream movers. Their influence propagates toward firms such as Stericycle (SRCL) and Codexis (CDXS), forming a layered but non-centralized structure.

Capital Goods

A small group of technology-oriented firms drives leadership. II-VI (IIVI) and Mercury Systems (MRCY) initiate movements that are followed by Nordson (MKSI) and FLIR Systems (FLIR), producing clear and directional influence paths.

Consumer Durables

Leadership is fragmented across independent channels. American Superconductor (AMSC), Central Garden & Pet (CENT), and iRobot (IRBT) each lead specific followers, resulting in localized influence rather than a single sector-wide hierarchy.

Consumer Non-Durables

The sector is characterized by a few strong but isolated links. Fossil Group (FOSL) leads Crocs (CROX), while Columbia Sportswear (COLM) influences Sanderson Farms (SAFM), with limited interaction beyond these pairs.

Consumer Services

Leadership is highly concentrated. DISH Network (DISH) emerges as the sole leader, with its movements preceding those of Starbucks (SBUX), indicating a narrow but detectable information channel.

Energy

Energy exhibits one of the strongest leadership structures. Firms such as TUSK Energy (TUSK), Diamondback Energy (FANG), and Viper Energy (VNOM) influence multiple downstream companies, including Centennial Resource Development (CDEV), reflecting strong and widespread propagation effects.

Finance

Leadership is distributed among several financial institutions. Carlyle Group (CG), SEI Investments (SEIC), and Principal Financial (PFG) influence firms such as Ameritrade (AMTD) and T. Rowe Price (TROW), forming a sparse but multi-path network.

Health Care

Health Care displays rich internal structure. Biomarín (BMRN), Intuitive Surgical (ISRG), and Gilead Sciences (GILD) act as prominent leaders, influencing firms such as Vertex Pharmaceuticals (VRTX) and Hologic (HOLX) through both positive and negative effects.

Miscellaneous

Leadership is spread across platform-oriented firms. HealthEquity (HQY), CoStar Group (CSGP), and MercadoLibre (MELI) influence several followers, including Zillow Group (ZG) and GoPro (GPRO), producing moderate but consistent directional links.

Public Utilities

The sector shows multiple localized leadership relationships rather than a dominant hub. Firms such as NextEra Energy (NEXT), ADTRAN (ADTN), and Spark Energy (SPKE) influence utilities like Clean Energy Fuels (CLNE) and Casella Waste Systems (CWST).

Technology

Leadership is driven by large platform and semiconductor firms. Broadcom (AVGO) leads both Apple (AAPL) and Microsoft (MSFT), while Meta Platforms (FB) and Adobe (ADBE) form an opposing influence channel, highlighting delayed information transmission.

Transportation

Transportation exhibits multi-channel leadership across logistics and airlines. Old Dominion Freight Line (ODFL), Hub Group (HUBG), and JetBlue (JBLU) influence firms such as Ryder (RYAAY) and Werner Enterprises (WERN), forming a well-connected but non-centralized network.

Takeaway

Across sectors, leadership effects are directional, heterogeneous, and concentrated. A limited set of firms consistently act as short-term influencers, while most stocks respond passively or exhibit no detectable leadership dynamics.

---

ETFs vs Equal Weighted Portfolios

---

What does ETF analysis reveal about the performance of market leaders,sector averages, and followers, and about overall market concentration? How does survivorship bias impact the validity of our conclusions, and what steps can be taken to reduce its effects? Which time window should be analyzed?

Introduction

This analysis explores the long-term performance dynamics between two distinct investment approaches within the NASDAQ market: Value-Weighted (ETFs) and Equally-Weighted Portfolios.

Beyond a simple performance comparison, this study serves as a proxy for understanding market structure:

We apply this comparative framework at two levels:

Key Question: Is market performance primarily driven by the concentrated power of a few “Leaders”, or does the diversified strength of the “Followers” provide a superior edge?

---

Methodology

---

1. Conceptual Framework: Leaders vs. Followers

---

2. Data Sources & Preparation

---

3. Portfolio Construction

---

4. Limitations: Survivorship Bias

---

5. Global & Sector Benchmarks

---

6. Performance Metrics

---

Global Market Analysis

We begin with an overview of the entire NASDAQ market, comparing the Fidelity Nasdaq Composite Index Tracking Stock (ONEQ) against our Equally-Weighted Global Portfolio (composed of 2,074 stocks).

Use the dropdown menu below to explore how this relationship changes over different time horizons:

Temporal Dynamics: The 2015 Regime Shift

To understand exactly when and why market leadership changed, we combined a discrete time-period breakdown with a rolling window analysis. Both methods point to a structural pivot in the middle of the last decade.

1. Discrete Period Analysis

We first broke down returns into four distinct market phases. The interactive visualizations below highlight the dramatic reversal in fortune across different time periods:

Period	ETF Return	EW Return	Difference	Market Context
2005-2008	-6.89%	0.77%	+7.66%	Pre-crisis period
2009-2012	17.45%	32.56%	+15.11%	Post-crisis recovery
2013-2016	15.89%	19.62%	+3.74%	Mid-bull market
2017-2020	10.90%	-3.99%	-14.89%	Late bull & COVID

2. Rolling Window Visualization

The reversal is even more apparent when viewed dynamically. The chart below shows the rolling 2-year return difference, pinpointing the “Regime Shift” specifically to the 2014-2016 period.

Rolling 2-year returns comparison. Green areas indicate periods where the Equally-Weighted portfolio (Followers) outperformed.

The “Pivot Point” (2014-2016):

• The “Breadth” Era (2009–2014): Post-crisis recovery was broad-based. Small and mid-cap companies (“Followers”) grew faster than established giants. This was a “stock picker’s market.”
• The “Concentration” Era (2015–2020): A clear regime shift occurred around 2015. Investors flocked to “Safe Haven” mega-caps (GAFAM), causing the standard ETF to outperform significantly while the average stock lagged behind.

---

Sector-by-Sector Analysis

Health Care Sector (IBB)

Performance Highlights:

• Equally-Weighted Return: 15.99% annually
• ETF Return: 9.95% annually
• Outperformance: +6.04% in favor of EW portfolio
• Sharpe Ratio: EW (0.78) vs ETF (0.42) - 86% better risk-adjusted returns

Insights: The Health Care sector shows the strongest outperformance of equally-weighted portfolios. With 564 stocks analyzed, the diversification benefits are clear. The biotech and pharmaceutical industries have significant variation between large-cap leaders and smaller innovative companies, allowing equally-weighted portfolios to capture growth across the entire sector spectrum.

Key Observations:

• EW portfolio showed remarkable resilience during market downturns
• Consistent outperformance throughout the 15-year period
• Higher volatility in EW portfolio was more than compensated by superior returns

---

Finance Sector (FTXO)

Performance Highlights:

• Equally-Weighted Return: 0.98% annually
• ETF Return: -7.55% annually
• Outperformance: +8.53% in favor of EW portfolio
• Sharpe Ratio: EW (0.05) vs ETF (-0.27) - ETF had negative risk-adjusted returns

Insights: The Finance sector tells a dramatic story. The 2008 financial crisis devastated large financial institutions (heavily weighted in the ETF), while the equally-weighted portfolio, by diversifying across 484 financial stocks, managed to stay positive. This is a powerful demonstration of how equal weighting can protect against sector-specific crises that disproportionately affect market leaders.

Key Observations:

• Both portfolios struggled during 2008-2009 financial crisis
• EW portfolio recovered faster and maintained positive returns
• ETF remained below initial investment value for most of the period
• Largest outperformance difference among all sectors

---

Technology Sector (QTEC)

Performance Highlights:

• Equally-Weighted Return: 12.82% annually
• ETF Return: 11.43% annually
• Outperformance: +1.39% in favor of EW portfolio
• Sharpe Ratio: EW (0.59) vs ETF (0.47) - 26% better risk-adjusted returns

Insights: Technology shows the smallest outperformance gap, suggesting that large tech companies (represented in the ETF) have been strong performers. However, equally-weighted portfolios still edge out, likely due to capturing growth from mid-cap and emerging tech companies that aren’t heavily weighted in the ETF.

Key Observations:

• Both strategies performed well in the tech sector
• EW portfolio showed more consistent growth trajectory
• Technology sector demonstrated strong overall returns regardless of weighting method

---

Consumer Services Sector (FTXD)

Performance Highlights:

• ETF Return: -2.59% annually
• Equally-Weighted Return: -7.21% annually
• Underperformance: -4.62% (ETF outperformed)
• Sharpe Ratio: ETF (-0.13) vs EW (-0.35) - ETF had better risk-adjusted returns

Insights: Consumer Services is the only sector where ETFs outperformed. This suggests that in this sector, market leaders (large retailers, service companies) have been more resilient than the average company. The sector faced challenges from e-commerce disruption and changing consumer preferences, where established leaders with scale advantages performed better.

Key Observations:

• Both portfolios had negative returns during this period
• ETF’s value-weighting favored resilient large-cap companies
• Sector transformation (retail disruption) may have hurt smaller companies more
• Only sector where concentration in leaders was beneficial

---

Sector Overview & Key Findings

Key Findings:

Equally-Weighted portfolios outperformed in 3 out of 4 sectors (75%)

Sector	ETF Return	EW Return	Difference	Winner
Health Care	9.95%	15.99%	+6.04%	✅ EW
Finance	-7.55%	0.98%	+8.53%	✅ EW
Technology	11.43%	12.82%	+1.39%	✅ EW
Consumer Services	-2.59%	-7.21%	-4.62%	❌ ETF

---

Key Insights & Implications

1. Diversification Benefits

Equally-weighted portfolios capture returns across the entire sector spectrum, not just market leaders. This is particularly valuable in sectors with high innovation and growth potential among smaller companies.

2. Crisis Resilience

During market downturns (2008 financial crisis, 2020 COVID), equally-weighted portfolios showed superior resilience. Diversification across many companies provided protection when large-cap leaders were hit hardest.

3. Sector-Specific Dynamics

• Health Care & Finance: Strong EW outperformance suggests these sectors have significant value beyond just the largest companies
• Technology: Smaller outperformance gap indicates large tech companies are also strong performers
• Consumer Services: ETF outperformance suggests scale advantages matter more in this sector

4. Risk-Adjusted Returns

Equally-weighted portfolios generally showed better Sharpe ratios, indicating superior risk-adjusted performance. The additional volatility was more than compensated by higher returns.

5. Market Leadership Question

The results challenge the assumption that market leaders (value-weighted) always outperform. In most sectors, the “average” company (equally-weighted) delivered better returns, suggesting that:

• Market capitalization may not always reflect future growth potential
• Smaller companies can drive significant sector returns
• Diversification provides both return and risk benefits

---

Conclusion

This comprehensive analysis reveals that equally-weighted portfolios outperformed value-weighted ETFs in 75% of sectors analyzed over the 2005-2020 period. The findings suggest that:

1. Diversification pays off: Spreading investments equally across sector companies captures growth opportunities beyond just market leaders

2. Crisis protection: Equally-weighted portfolios showed superior resilience during market downturns

3. Sector matters: Performance varies significantly by sector, with Health Care and Finance showing the strongest EW advantages

4. Risk-adjusted superiority: Better Sharpe ratios indicate that EW portfolios provide superior risk-adjusted returns

5. Market leadership isn’t everything: Large market cap doesn’t always translate to superior returns

---

Global Conclusions: The Complete Market Picture

---