Why Standardized Earnings Surprises Are Not Automatically Z-Scores

Summary

The document asks whether a standardized unexpected earnings score can be read as a z-score when earnings surprises are assumed to follow a normal distribution. Its answer highlights a distinction between standardizing an observation and knowing the population standard deviation: the score’s standard deviation is estimated. On that basis, the response says the statistic may be closer to a t-score than a z-score, and rejects treating it as certainly a z-score.

This is a concise statistical caution relevant to interpreting earnings-surprise measures. The exchange gives no derivation, sample-size details, estimation procedure, or conditions under which a t-distribution approximation would apply. It also does not establish whether the underlying surprise data are normally distributed; normality is part of the question’s premise. Researchers should therefore treat the answer as a reminder about estimated scale parameters, not as a complete specification for inference or a validation of any particular SUE calculation.

Key ideas

• A standardized earnings surprise is not automatically a z-score merely because the underlying data are assumed normal.
• The response points out that the standard deviation used in the score is estimated.
• An estimated standard deviation can make a statistic more like a t-score, depending on the setup.
• The exchange does not provide enough detail to establish a specific distribution or inference procedure.

Tags

# Can Standardized unexpected earnings be considered a Z-score


# Can Standardized unexpected earnings be considered a Z-score












According to this wikipedia: http://en.wikipedia.org/wiki/Earnings_surprise, the SUE score is a "standardized" difference between reported earnings and expected earnings. Therefore, can the SUE score be interpreted as a Z-score assuming that the underlying earnings surprise data is normally distributed?

## Answer by James (score 0, accepted)

https://quant.stackexchange.com/a/14717

No, because the $\sigma$ is estimated, and therefore SUE may be a t-score, but certainly not a z-score.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.