The balance-sheet proxy has a precise meaning
An investment factor need not measure a company's capital expenditure directly. The French investment-portfolio documentation defines its sorting variable as the change in total assets between two fiscal years divided by the earlier year's assets, with portfolios formed in June. That observable accounting proxy includes more than expenditure on factories or software. Acquisitions, disposals and other balance-sheet changes can influence it.
Start with the research question: is the hypothesis about overall balance-sheet expansion, organic capital expenditure, or investment relative to operating cash generation? Each demands different fields. Replacing total-asset growth with capital expenditure divided by assets creates a new descriptor. It may be reasonable, but it is not a correction that can inherit another factor's evidence without an independent evaluation.
Separate growth rates from their causes
Suppose fictional Firm D moves from $1,000 million to $1,180 million of total assets. Its asset-growth rate is 18%. A simplified acquisition bridge attributes $150 million of the increase to purchased assets and $30 million to all other net changes. Firm E moves from $500 million to $550 million, producing 10% growth without an acquisition in this illustration.
D has the higher raw growth rate, but its non-acquisition increase relative to beginning assets is only 3%. That supplementary bridge answers a different question from the 18% descriptor. Preserve both observations and their definitions. Do not remove acquisitions only in periods when they hurt results; a permanent adjustment rule requires reliable historical transaction information, accounting treatment and publication timing.
Handle comparable periods and exceptional denominators
Check that the two observations represent comparable fiscal-year boundaries. A changed year-end, restatement or reporting-currency change can make an apparent annual comparison span another interval. Keep original report identifiers and effective dates. If a vendor overwrites prior assets, rerunning an old portfolio with the revised denominator changes the historical signal even when the current-period numerator is unchanged.
Very small beginning assets can create extreme ratios, especially after a restructuring or newly public history. Excluding those records, imposing a minimum reporting history, or clipping the score can all be defensible if declared before results. Treat an unavailable earlier balance sheet as missing, not zero. Record an eligibility reason so later reviewers can distinguish insufficient data from an intentional economic exclusion.
Interpret factor exposure with restraint
A lower-growth portfolio is not necessarily safer, more profitable or better governed. Shrinking assets can reflect discipline, distress or a disposal. High growth can reflect productive opportunities or expensive expansion. An empirical association between a descriptor and returns does not identify the mechanism for a particular company. Keep any causal interpretation separate from the arithmetic and from the portfolio's measured exposures.
Useful checks include comparing raw growth with a predeclared acquisition-aware variant, inspecting sector composition, and measuring overlap with value and profitability. Publish every planned comparison, including failures. Apply the same timing and cost assumptions to each. This worked example contains no observations from real issuers and cannot establish an investment premium; its purpose is to show how an apparently simple factor can conceal several economic stories.
Inspect the example
| Fictional firm | Beginning assets | Ending assets | Raw growth | Other net increase / beginning assets |
|---|---|---|---|---|
| D | $1,000m | $1,180m | 18% | $30m / $1,000m = 3% |
| E | $500m | $550m | 10% | $50m / $500m = 10% |
Sources & reproduction
Primary references support the definitions and method. Numerical examples are illustrative unless a source record is explicitly identified.