A z score of -3 means a value is three standard deviations below the mean of its reference distribution. If that distribution is normal, the value is unusually low: about 0.13% of observations fall below it, and about 0.27% lie at least three standard deviations from the mean in either direction.
What a z score tells you
A z score, also called a standard score, expresses a value’s position relative to a mean in units of standard deviation. It makes measurements from different scales easier to compare, provided you know which distribution and reference group are being used.
The standard formula is z = (x − μ) / σ, where x is the observed value, μ is the population mean, and σ is the population standard deviation. For sample data, analysts often use the sample mean and sample standard deviation instead. A score of -3 says the observed value is three standard deviations less than the mean; it does not say the value is three units below it.
- A negative z score is below the mean.
- A positive z score is above the mean.
- A z score of 0 is exactly at the mean.
How to calculate a score of -3
To find the value that corresponds to a z score of -3, rearrange the formula: x = μ + zσ. Since z is -3, this becomes x = μ − 3σ.
For example, suppose a test has a mean of 100 and a standard deviation of 15. A score with z = -3 is 100 − (3 × 15), or 55. In another measurement with a mean of 50 and standard deviation of 4, the same z score corresponds to 38. The standardized position is the same, even though the original values and units differ.
Use the standard deviation that matches the data and reference group. If a report gives a z score directly, check its documentation for the population, age group, test version, and calculation method; those details affect what the comparison means.
How unusual is -3?
For a perfectly normal, or bell-shaped, distribution, a z score of -3 is at the third standard deviation below the mean. The cumulative probability to its left is about 0.00135, or 0.135%—roughly 1 to 740 observations. The probability between -3 and +3 is about 99.73%, a result often summarized by the empirical rule.
These probabilities rely on the normal-distribution assumption. A z score alone does not guarantee that a value is rare. In a skewed distribution, a distribution with heavy tails, or a small dataset, the proportion below -3 may be substantially different. If the data are not approximately normal, use the observed data or an appropriate model to estimate how unusual the value is.
Also distinguish “at -3” from “-3 or lower.” For a continuous normal distribution, the probability of one exact value is effectively zero; the useful probability is that of a range or tail, such as values at or below -3.
What it does—and does not—mean
A z score is a relative position, not a verdict. It does not by itself show why a value is low, whether it is harmful, or whether an error occurred. Interpretation depends on the measurement, the quality of the data, and the purpose of the comparison.
- It is not a percentage. A z score of -3 does not mean a score is 3% lower than average.
- It is not three standard deviations as a percentage of the value. Standard deviations measure spread around the mean, not a percentage change.
- It does not prove a result is wrong. Rare observations can be genuine; unusually low results can also arise from measurement or data-entry errors.
- It is not automatically a diagnosis or a measure of risk. In health or education, use the relevant professional guidance and the specific reference standards.
When a value appears unexpectedly low, check the units, the mean and standard deviation used, and whether the correct comparison group was selected. Consider whether the underlying distribution is normal and whether the measurement was collected reliably.
Comparing -3 with other z scores
Under a standard normal model, more negative scores are farther below the mean and generally have smaller lower-tail probabilities. For context, about 2.3% of values fall below -2, compared with about 0.135% below -3. A score of -1 is much less unusual: about 15.9% fall below it.
A score of +3 has the same distance from the mean as -3, but on the high side. In a symmetric normal distribution, the two tails beyond -3 and +3 each contain about 0.135%. The combined probability of being at least three standard deviations away in either direction is about 0.27%.
These are model-based reference figures, not universal cutoffs for significance or importance. In statistical testing, a z score’s meaning depends on the test, whether the question is one-sided or two-sided, and the chosen decision threshold. A rare observation is not automatically statistically significant evidence for a particular claim.
Frequently Asked Questions
Is a z score of -3 good or bad?
It means the value is three standard deviations below its reference mean. Whether that is good, bad, or concerning depends on what is measured and the relevant standards; the score alone does not supply that judgment.
What percentile is a z score of -3?
In a standard normal distribution, -3 is approximately at the 0.13th percentile: about 0.13% of values are lower. The percentile may differ if the actual distribution is not normal.
What raw value corresponds to z = -3?
It is the mean minus three standard deviations: x = μ − 3σ. You need the mean and standard deviation to calculate a particular raw value.
Is -3 statistically significant?
Not necessarily. A z score of -3 can be associated with a small tail probability under a normal model, but statistical significance depends on the analysis, hypothesis, tail direction, assumptions, and threshold.
Can a z score be less than -3?
Yes. Z scores have no fixed lower limit. A score below -3 is farther below the mean, though its rarity must be judged using the distribution and reference data.
How to interpret it responsibly
Read -3 as “three standard deviations below this particular mean,” then ask what population and distribution define that comparison. If the data are approximately normal, the result lies in a very low tail; if not, the normal-based probability may mislead. For practical decisions, combine the score with the original measurement, data quality, context, and any field-specific guidance.
