in building a fundamental database for equity investors - how do i calculate EPS growth,…
- User
in building a fundamental database for equity investors - how do i calculate EPS growth, knowing that some companies have negative eps in the past, some have negative in the future etc. - ie. growth from positive to positive is clear, but negative to negative, negative to positive is quite a problem
- Assistant
Great question! Calculating EPS growth when EPS values can be negative or cross zero is tricky because the usual percentage growth formula:
[ \text{EPS Growth} = \frac{\text{EPS}{t} - \text{EPS}{t-1}}{|\text{EPS}_{t-1}|} ]
can produce misleading or undefined results (division by zero, or nonsensical large negative growth).
Challenges with EPS Growth Calculation
- Positive to positive EPS: straightforward percentage growth.
- Negative to negative EPS: percentage growth is ambiguous (e.g., from -2, is that growth or decline?).
- Negative to positive EPS: growth from a loss to profit — percentage growth is not meaningful (division by negative number).
- Positive to negative EPS: similarly problematic.
Recommended Approaches for EPS Growth in a Fundamental Database
1. Use Absolute Change Instead of Percentage Growth
Calculate the absolute change in EPS:
[ \Delta \text{EPS} = \text{EPS}{t} - \text{EPS}{t-1} ]
- This avoids division by zero or negative base issues.
- You can track whether EPS improved or deteriorated by the sign of the change.
- Useful for flagging trends without misleading percentages.
2. Use a Modified Growth Metric with Sign Handling
Define growth as:
[ \text{EPS Growth} = \frac{\text{EPS}{t} - \text{EPS}{t-1}}{\max(|\text{EPS}_{t-1}|, \epsilon)} ]
where (\epsilon) is a small positive number (e.g., 0.01) to avoid division by zero.
- This caps extreme growth rates when EPS is near zero.
- Still problematic if EPS changes sign, but less volatile.
3. Use Logarithmic Growth for Positive EPS Only
For positive EPS values only, use log growth:
[ \text{Log EPS Growth} = \ln(\text{EPS}{t}) - \ln(\text{EPS}{t-1}) ]
- This is symmetric and scale-invariant.
- But only valid if both EPS values are positive.
- For negative or zero EPS, this is undefined.
4. Categorize EPS Transitions
Create a categorical variable to capture EPS sign transitions:
Previous EPS Current EPS Category Positive Positive Positive to Positive Positive Negative Positive to Negative Negative Positive Negative to Positive Negative Negative Negative to Negative Zero Any Zero to Any Any Zero Any to Zero - Use this to flag and treat growth calculations differently.
- For example, report absolute change or mark as "transition" cases.
5. Use Median or Average Growth Over Multiple Periods
- Calculate growth over multiple years to smooth volatility.
- For example, use CAGR (Compound Annual Growth Rate) over 3-5 years, but only if EPS remains positive throughout.
- If EPS crosses zero, report "N/A" or use absolute change.
Summary Table of EPS Growth Calculation Methods
Method Pros Cons When to Use Percentage Growth Intuitive for positive EPS Misleading or undefined with negatives or zero Positive to positive EPS only Absolute Change Simple, no division issues Not scale-invariant All cases Modified Growth with Epsilon Avoids division by zero Still tricky with sign changes When EPS near zero Logarithmic Growth Symmetric, scale-invariant Only for positive EPS Positive EPS only Categorical Transition Flags Helps identify problematic cases Requires manual handling For data quality and analysis
Practical Implementation Tips
- Store raw EPS values and store growth metrics separately.
- For growth metrics, flag cases where EPS crosses zero.
- Consider reporting multiple growth metrics (absolute change + percentage growth where valid).
- Document your methodology clearly for users of the database.
If you want, I can help you with a sample Python or Excel formula to implement these approaches! Would you like that?
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.