Style Box Analysis (Morningstar 3×3)

The Morningstar Style Box is a 3×3 grid that classifies equity portfolios on two axes: size (large / mid / small market cap) and value-growth orientation. FM103 computes the style-box position of each rebalance period's portfolio and tracks the cell occupancy over time. Style drift is often the first visible sign that a strategy is doing something different from what its name suggests.

The two axes

The original Morningstar methodology (Morningstar 1992) classifies each holding by:

• Size: rank by market cap within the universe. Top 70% by cumulative market cap is "large," next 20% is "mid," bottom 10% is "small." (The split is by cap percentile, not stock count.)
• Value-growth: composite of price-multiple metrics. Names with low P/E, P/B, P/CF are "value"; high are "growth"; middle is "blend."

FM103 follows the same convention with FMP factor data: market cap for the size axis, blended P/E + P/B + P/CF z-score for the value axis.

Per-period cell occupancy

For each rebalance period the sub-pill computes the share of portfolio weight in each of the 9 cells:

• Large Value, Large Blend, Large Growth
• Mid Value, Mid Blend, Mid Growth
• Small Value, Small Blend, Small Growth

The dominant cell is highlighted. A strategy whose dominant cell stays the same across all periods is style-consistent. A strategy whose dominant cell shifts is style-drifting.

Style drift score

The aggregate drift metric is the average L1 distance between consecutive period style vectors:

where w_cell,t is the portfolio weight share in cell at period t. The metric ranges from 0 (no drift) to 2 (complete rotation period-to-period). Thresholds:

• Drift < 0.30: Style-consistent. Good.
• Drift 0.30–0.60: Moderate rotation; check whether intentional.
• Drift > 0.60: Strategy is essentially style-rotating. Consider whether the factor specification needs explicit style constraints.

Why drift matters

Style drift hides factor exposure changes that aren't obvious from the factor weights. A "value + momentum" strategy with 50/50 weighting can drift from large-value (in flat markets) to small-growth (in momentum-driven markets) without changing any configuration. The drift signature reveals this.

Drift also affects attribution. Factor attribution (see Factor Attribution) is computed against the named factors. Drift into a style cell whose returns differ from the named factors' returns shows up as elevated specific return — the model can't name the source.

The 3×3 timeline visualisation

The sub-pill shows a per-cell stacked-area chart over time: x-axis is rebalance periods, y-axis is cumulative cell share. The 9 stack colors immediately reveal whether the strategy is layered (constant cell mix), rotating (one cell dominates then another), or drifting (gradual shift in one direction).

The dominant-cell label

Most backtests aggregate to one dominant cell ("Large Blend" for a market-cap-weighted S&P factor strategy). This label, paired with the cumulative cell distribution, lets the user describe the strategy honestly. A strategy labelled "Quantitative Value" that style-boxes to Large Blend has a naming problem — investigate whether the value tilt is being washed out by mechanical equal-weighting in the top-N selection.

Limitations

• The size cutoffs are universe-dependent. A "large" cap in a small-cap universe is still small in absolute terms.
• The value-growth axis depends on what metrics are blended. Different choices produce different classifications for marginal names.
• The methodology doesn't cover non-equity allocations — cash, fixed income, alternatives all sit outside the box.