- Strategy: Consider a "Quantamental" Approach – Blending Quantitative and Fundamental Analysis. Causeway's core philosophy is that a purely quantitative ("quant") approach excels in breadth (covering many stocks systematically) and consistency, while fundamental analysis provides in-depth understanding of individual companies. Combining these can leverage the strengths of both. [04:15-05:38]
- Method – Fundamental Risk Overlay: They employ fundamental analysts to act as a "sanity check" or a "fundamental layer of risk control" on quant model outputs. This is crucial for identifying risks quant data might miss, such as the binary event risk of an upcoming clinical trial for a pharmaceutical company, regulatory reviews, or significant litigation. This prevents investing in stocks that look attractive quantitatively but face imminent, non-quantifiable threats. [06:01-07:39]
-
Factor Investing Methodology: Employ a "Committee" Approach. Instead of relying on a single definition for a factor, or isolated factor portfolios, Causeway views their seven factor categories (Value, Sentiment, Growth, Quality, Technicals, Corporate Events, Sustainability) as different "committees," each "voting" on a stock's attractiveness. A stock doesn't need unanimous approval to be included. [00:00-00:08, 36:01-36:12]
-
Factor - Value: Sophisticated Definition Beyond Simple Metrics.
-
Method – Evolution of Value: They acknowledge that simple Price-to-Book (P/B) is less effective now. [12:31-12:33]
-
Method – "Warranted Multiples": They refine P/B by considering what multiple is "warranted" given other company characteristics. For example, a company with a high Return on Equity (ROE) can "support" or justify a higher P/B multiple. This implicitly accounts for intangible assets that might drive high ROE. [12:35-13:05]
-
Method – Composite Value Inputs: They use a range of valuation measures: cash flow yields, "payout yield" (dividends + buybacks), earnings yields, and EV/EBITDA. A stock appearing cheap across multiple metrics increases confidence. [13:31-13:53]
-
Method – Incorporating Sell-Side Target Prices: Unconventionally, they include consensus sell-side price targets as a value input. The rationale is that analysts' discounted cash flow (DCF) models often look further out than typical 1-2 year earnings estimates, potentially capturing longer-duration value, especially when near-term earnings are disrupted (e.g., by COVID). [13:56-15:02]
- Factor - Sentiment: Predictive of Near-Term Earnings Momentum. This factor aims to predict earnings momentum over the next 3-6 months.
-
Method – Inputs:
-
Sell-side analyst estimate revisions and estimate diffusion (the breadth of revisions). [15:55-16:00]
-
Earnings announcement returns (market reaction to earnings). [16:03-16:05]
-
Natural Language Processing (NLP) on Earnings Call Transcripts: They've developed a proprietary system to assess the sentiment expressed by management during earnings calls using NLP techniques. [16:07-16:24]
-
"Analyst Relative Ideas": A measure to determine which stocks are broadly favored or disfavored by the sell-side community, reflecting behavioral incentives and buy-side discussions. [16:26-16:53]
-
- Factor - Technicals/Momentum: Combining Traditional and Network-Based Approaches.
-
Method – Traditional Momentum: Includes standard 6, 9, and 12-month price momentum. [17:25-17:37]
-
Method – "Network Momentum" (or Linkage Momentum): This is a more unique aspect. Instead of just looking at a target stock's momentum, they identify its "peers" and measure the weighted average momentum of those peers. The theory is that market information diffuses into stocks at different rates; larger, well-covered stocks react quickly, while news might take longer to impact related (e.g., competitor, supply chain) companies. Network momentum aims to capture this leading/lagging effect. [00:27-00:42, 17:37-18:55]
-
Method – Defining Peers for Network Momentum: Peers are identified through various means: direct competitors, customers, companies covered by the same sell-side analysts (who tend to group them logically), and an NLP-based method analyzing earnings call transcripts to find companies discussing similar challenges, opportunities, or technologies. [19:20-20:01] (Example: Clear Secure [20:04-20:55])
- Factor - Growth: A Nuanced Approach, Not Just the Opposite of Value.
- Method – Interaction with Value: Growth is used to identify companies that might "deserve" higher valuation multiples if their growth characteristics are strong. The model looks for stocks where the market may not have fully appreciated the growth potential relative to its current valuation. This is a form of "warranted multiple" thinking. [20:58-22:35]
- Factor - Quality (Internally termed "Competitive Strength"):
-
Method – Multi-faceted Inputs: Assesses the attractiveness of the industry, the company's market share and defensibility within that industry, profit margins, and overall profitability. It also incorporates balance sheet and financial strength metrics. [23:36-24:23]
-
Method – Emphasis on Balance Sheet for Small Caps: Recognizes that small-cap companies may have less access to capital, making balance sheet strength particularly crucial. [24:49-25:01]
- Factor - Corporate Events: Systematically Identifying Event-Driven Mispricings. This factor aims to capture the market's typical (often drifting) reaction to specific corporate events.
-
Method – Event Studies & Post-Announcement Drift: They conduct event studies on hundreds of corporate event types to determine if there's a statistically significant "post-announcement drift" (i.e., does the stock price tend to drift in a certain direction for months after the event?). [27:31-28:07]
-
Method – Stacking Low-Breadth Events: Many individual events are infrequent ("low breadth"). By identifying ~20 events with predictable drifts and giving each a coefficient (weight) based on the strength of that drift, they stack these to create a composite "Corporate Events" factor with good breadth. [28:07-28:37]
-
Method – Contextualizing Event Impact (Refinement): They recognized that the same event (e.g., an equity issuance) can have different implications depending on the company's situation. A distressed company issuing equity to fix its balance sheet might see a positive market reaction, while a healthy company diluting shareholders might see a negative one. To account for this, they observe the 3-day market return around the event announcement . If the market reacts contrary to their prior (e.g., stock rallies on an equity issuance), the weight of that specific event for that stock is revised, potentially down to zero. This adds a layer of market-based validation. [28:51-30:20]
- Factor - Sustainability (ESG): Focused on Materiality and Alpha Generation.
-
Method – Materiality is Key: They fine-tune ESG inputs down to those they believe are material to investment performance, rather than a broad "feel-good" approach. The goal is to build a sustainability factor that "makes people money." [00:50-00:57, 32:50-33:24]
-
Method – Governance (G) as a Core Component: Good governance is considered uncontroversial and crucial; a "garbage management team" can ruin an otherwise attractive investment. [33:26-33:48]
-
Method – Relative Alpha Contribution: While valuable, the sustainability factor is seen as a "second-tier" alpha contributor compared to major drivers like Value and Sentiment. It might add an incremental 20-30 basis points of alpha per year. [34:00-34:31]
- Portfolio Construction: Composite Scores with Contextual Weighting.
-
Method – Composite Factor Score: They prefer blending all seven factor category scores into a single composite score for each stock, rather than managing separate factor "sleeves." This allows for interactions between factors. [35:40-37:32]
-
Method – "Contextual Weighting" of Factors: This is a distinct feature. The weighting of the seven main factor categories (Value, Sentiment, etc.) in a stock's final score is dynamically adjusted based on the stock's own characteristics.
-
Rationale: "Clientele Effects." Different types of investors (value, growth, etc.) prioritize different factors.
-
Process: Stocks are first classified into broad "dimensions" (e.g., appears to be a value stock, a growth stock, or a core/blend stock) based on how they screen to these different investor clienteles. Then, the weights of the seven alpha factors are re-estimated specifically within these cohorts. For example, for stocks that screen as "high value," the value factor itself might get a higher weight in their final composite score, while for "high growth" stocks, growth and sentiment factors might get higher weights. This allows the model to be "more receptive" to a stock like a Mag 7 company that looks expensive on pure value but has a great growth story. [00:15-00:25, 39:21-44:40]
-
- Risk Management: Proprietary, Customized Risk Models.
-
Method – In-House Development: Causeway builds its own risk models rather than relying solely on commercial ones (e.g., Barra, Axiom). This allows them to incorporate factors specific to their investment process and alpha factors. [45:48-46:09]
-
Method – Custom Risk Factors: They include a "cyclicality" risk factor, for instance, which is important if a value-driven process leads to more cyclical stocks. [46:09-46:31]
-
Method – Universe-Specific Estimation: Risk models are estimated separately for different universes (e.g., Emerging Markets, International Small Cap) because factor behavior (e.g., value or growth risk premia) can differ significantly across these segments. A global, one-size-fits-all risk model might miss these nuances. [46:36-47:12]
-
Method – Factor Categories in Risk Model: Include style factors (value, growth, momentum, cyclicality, size, volatility), as well as geographic, currency, and sector/industry factors. [47:24-47:51]
-
Method – Forward-Looking Risk Prediction: Risk is assessed based on a portfolio's current exposures to these systematic factors and the historical covariance matrix (volatilities and correlations) of these factors. [47:53-48:33]
-
Key Risk Factor: The "volatility factor" (looking at stock betas, recent return volatility, high-low trading range) often explains the largest amount of risk. [48:52-49:16]
- Portfolio Sizing & Concentration: A Balance, Leaning Towards Concentration.
-
Method: They typically hold a smaller number of stocks (e.g., ~150 in a 6,000-stock global universe) than many quant funds. This allows them to focus on the highest conviction names (far out on the right tail of expected returns). [49:41-51:18]
-
Justification: They believe that with 150 stocks, they can still achieve sufficient diversification across countries, sectors, and risk factors. The fundamental risk control overlay (analyst sanity checks) gives them more comfort in running somewhat more concentrated portfolios than a pure quant firm might. [51:18-51:57]
- Rebalancing & Sell Discipline: Dynamic and Opportunistic.
-
Method – Daily Optimization: The model runs optimizations every day to identify the ideal portfolio. [53:18-53:23]
-
Method – Variable Turnover Based on Alpha per Unit of Turnover: Instead of fixed rebalancing dates (e.g., monthly), they assess daily how much alpha (expected return) increase can be achieved per unit of turnover (trading). They run optimizations at various turnover levels (e.g., 2%, 4%, 8%, 16%, 20% of the portfolio). They trade more when this ratio is high (i.e., many good opportunities exist, like during market dislocations such as "Liberation Day") and trade less (or not at all) when the potential alpha gain doesn't justify the trading costs and market impact. The average annual turnover might be, for example, 95%, but it's applied opportunistically. [53:23-55:59]
-
Method – Sell Discipline: A stock is sold if it "can't compete for its place in that portfolio anymore" based on the optimizer's assessment of its expected return versus its risk characteristics. [52:53-53:16]
- Application of Machine Learning (AI): Identifying Non-Linear Relationships and Interactions.
-
Method – Tree-Based Models (XG Boost, Random Forest): Used specifically within the Quality factor to capture non-linearities and interactions between different quality sub-factors. For example, a stock with high net debt and poor financial strength (a cash flow measure) might be penalized more severely by the ML model than a linear model would suggest, reflecting a synergistic negative effect. [56:30-58:38]
-
Method – Emphasis on Transparency & Interpretability: Despite using ML, they invest in visualization tools and methods to understand why the ML model makes certain decisions (e.g., how much each input feature contributes to the output score). They avoid a "black box" approach, ensuring portfolio managers can comprehend and trust the model's reasoning. [58:45-59:33]
- Key Lesson for Investors: Embrace Humility and Robust Risk Management.
-
Forecasting is inherently difficult; even the best will have a hit rate likely in the 60-70% range. [1:05:20-1:05:43]
-
Therefore, diligent risk management is paramount to ensure that skill can manifest over time and to avoid being overwhelmed by inevitable forecasting errors or periods of factor underperformance. [1:06:18-1:06:35]
- Warning: Don't Abandon Well-Established Factors Prematurely.
-
Factors rooted in persistent human behavioral biases (like Value, which exploits overreaction and underreaction) are likely to remain effective over the long term, even if they experience periods of underperformance. [1:03:43-1:05:11]
-
Avoid being swayed by short-term market "moods" or "groupthink" to discard a sound, long-term factor-based strategy. [1:03:20-1:03:26, 1:04:25-1:04:39]