What Is Kurtosis? How to Read Kurtosis in SPSS
What Is Kurtosis
You open your SPSS file, go to the Descriptive Statistics table, and see a Kurtosis column for each variable. This statistic describes how peaked or flat the data distribution is compared with a normal distribution. More specifically, kurtosis reflects how concentrated the observations are around the center and how heavy the tails of the distribution are.
In quantitative research, SPSS usually displays Kurtosis together with Std. Error of Kurtosis. Kurtosis can be positive, negative, or close to 0. With the calculation commonly used in SPSS, the kurtosis of a normal distribution is close to 0. You should therefore read kurtosis in the context of the distribution checks, sample size, and analysis you plan to run, rather than looking at the number on its own.
Conceptually, kurtosis is calculated from the deviations of individual observations from the mean and then standardized using the standard deviation. You do not need to calculate it by hand when your data are already in SPSS. The more useful question is what the statistic is telling you: whether the data are too peaked, too flat, or have unusually heavy tails that require further checking.
Kurtosis is related to, but different from, standard deviation. Standard deviation measures how spread out the values are around the mean, while kurtosis describes the shape of the distribution, especially its concentration and tails. If you are reviewing how dispersed your data are, you can also read about variance before interpreting kurtosis.
The Meaning of Kurtosis in Quantitative Research
Kurtosis is commonly used during exploratory data analysis and assumption checking. After entering your questionnaire data into SPSS, you can inspect kurtosis for each item or for the composite score of each scale. The statistic indicates whether the data have a shape that differs substantially from a normal distribution.
Positive kurtosis indicates a distribution that tends to be more peaked, with heavier tails than a normal distribution. Many observations may cluster around a particular range while some observations remain far from the center. Negative kurtosis indicates a flatter distribution, with observations spread more broadly around the center. Kurtosis close to 0 indicates a shape close to normal according to this measure.
Use kurtosis to decide what to check next. Do not use it as an automatic reason to delete an item. If a variable has high kurtosis, inspect its histogram, boxplot, minimum, maximum, and unusual cases. A single statistic cannot establish that the data are wrong or that the questionnaire has a problem.
Kurtosis can matter in analyses that involve distributional assumptions, such as linear regression, t-tests, or ANOVA. The effect also depends on sample size, sampling method, how balanced the data are, and the structure of the model. For analyses based on Likert scales, checking kurtosis is useful for diagnosis, but you still need to read it alongside the other checks.
Distinguish the kurtosis of individual variables from the kurtosis of regression residuals. Before running the model, you can check the distribution of an item or a composite variable. After running the regression, the residuals are the object you examine when assessing the model assumptions. These are two different purposes.
You can also review the articles in the Statistics topic if you are checking mean, median, variance, and standard deviation at the same time.
What Level of Kurtosis Is Acceptable
The question “what level of kurtosis is good” does not have one number that applies to every research project. First identify whether you are using kurtosis to describe the data, assess approximate normality, or detect a variable that needs review. Some practical references use absolute values of skewness and kurtosis for an initial assessment, but the selected cutoff must be stated with its source and should not be treated as a fixed rule for every sample size.
The table below summarizes common interpretations. The rows describe what the statistic means. When you make a quantitative claim in your thesis, state the criterion and the methodological source you used.
| How to read kurtosis | Practical meaning | Source or note |
|---|---|---|
| Close to 0 | The distribution shape is close to a normal distribution according to the kurtosis statistic | The excess kurtosis calculation commonly used in SPSS |
| Positive kurtosis | The distribution tends to be more peaked, and the tails may be heavier | Check the histogram and outliers as well |
| Negative kurtosis | The distribution tends to be flatter | Check the sample size and the actual shape as well |
| A higher absolute value | The distribution departs further from normality and may contain unusual observations | Combine checks of skewness, boxplot, and extreme values |
| Using a cutoff in the thesis | State the cutoff, calculation method, and reference clearly | You can refer to the SPSS diagnostic procedure in (Field, 2013) |
A more structured check is to calculate a standardized value by dividing Kurtosis by Std. Error of Kurtosis. If the absolute value of the result is large, the data may depart substantially from normality according to this approximate test. Be careful when interpreting the result with a large sample. A large sample can make a small difference in shape statistically significant even when its practical effect on the analysis is limited.
State the criterion you used in your thesis. If your supervisor requires a specific range, use the agreed criterion and cite the corresponding source. Do not take a number from a website and present it as a universal rule. Methodological references such as (Field, 2013) can support your explanation of data checking and diagnosis in SPSS, but you still need to describe the table and calculation method you actually used.
How to Read Kurtosis in SPSS Output
To obtain kurtosis in SPSS, go to Analyze > Descriptive Statistics > Descriptives. Move the variables you want to check into Variable(s), select Options, tick Mean, Std. deviation, Minimum, Maximum, Skewness, and Kurtosis, then select Continue and OK. SPSS will generate the Descriptive Statistics table.
To inspect a chart, go to Analyze > Descriptive Statistics > Explore. Move the variable into Dependent List, open Plots, select Histogram, and optionally select Normality plots with tests. This gives you a histogram, boxplot, and distribution tests. When reading the output, start with the variable name and sample size N, then move to Skewness and Kurtosis.
Illustrative output
The table below is illustrative output, not the result of an actual study. The column names follow the Descriptive Statistics table printed by SPSS.
| Variable | N | Minimum | Maximum | Mean | Std. Deviation | Skewness | Kurtosis |
|---|---|---|---|---|---|---|---|
| HL1 | 220 | 1.00 | 5.00 | 3.74 | 0.86 | -0.48 | 0.21 |
| HL2 | 220 | 1.00 | 5.00 | 3.61 | 0.91 | -0.31 | -0.44 |
| HL3 | 220 | 1.00 | 5.00 | 3.69 | 0.88 | -0.62 | 1.18 |
| HL4 | 220 | 1.00 | 5.00 | 3.77 | 0.83 | -0.27 | 0.06 |
In this illustrative table, HL1 has kurtosis of 0.21, meaning that its distribution is slightly more peaked than normal but still fairly close to 0. HL2 has negative kurtosis, indicating a slightly flatter distribution. HL3 has a value of 1.18, higher than the other three variables, so you should inspect its histogram and boxplot as well.
If SPSS also displays Std. Error for skewness and kurtosis, you can calculate the standardized value. For example, if HL3 has Kurtosis = 1.18 and Std. Error of Kurtosis = 0.327, the ratio is approximately 3.61. This is a reason to investigate further, but it is not enough to conclude that HL3 must be deleted. Check the original data, scale coding, and extreme values.
Do not confuse Kurtosis with Std. Error of Kurtosis. The first column is the estimate of the distribution shape. The second represents the uncertainty of that estimate. When calculating the standardized ratio, divide Kurtosis by Std. Error of Kurtosis, not by Std. Deviation.
How to write this in your thesis
You can write the following: “The descriptive statistics results show that the kurtosis values of the items ranged from [lowest value] to [highest value]. Variable [variable code] had the highest kurtosis, so the research team continued to examine its histogram, boxplot, and extreme values before conducting [next analysis].” Replace every bracketed placeholder with the actual values from your file.
What to Do When Kurtosis Is Outside Your Criterion
When kurtosis has an unusually high or low value, work from data checking to methodological review. This gives you a defensible explanation for your supervisor and reduces the risk of deleting an item simply to make the results table look cleaner.
Check data entry and coding
Open Variable View and Data View, then check Measure, value labels, the permitted coding range, and reverse-coded variables. A Likert item coded from 1 to 5 that contains a value of 55 can sharply increase kurtosis. If a reverse-coded item has not been recoded, the distribution and correlations within the scale may also be affected.
Check outliers and unusual responses
Use Analyze > Descriptive Statistics > Explore, then inspect the boxplot and the list of extreme cases. You can read more about outliers to distinguish a data-entry error from a valid respondent with an unusual pattern. Delete an observation only when you have a clear methodological reason, such as invalid data or a violation of cleaning criteria established in advance.
Recheck the items
If one item has a markedly different kurtosis value from the other items in the same scale, review the item wording, number of response levels, and response percentage at each level. An item selected at the same response level by many respondents can produce a peaked distribution. Any deletion should also consider Corrected Item-Total Correlation, Cronbach's Alpha, EFA, or the measurement model, depending on your research procedure.
Consider transformation or another analysis method
For some projects, you may consider transforming the data, but explain clearly how the transformation changes the interpretation of the variable. With Likert data, automatically applying a log or square-root transformation is often not the first choice because it makes the results harder to explain.
If the distributional assumption is not appropriate, you may consider a method that is less sensitive to distributional departures or a suitable alternative test. SPSS, JASP, and R offer different options. SmartPLS also uses its own estimation approach, but you should not take a distribution cutoff from CB-SEM and apply it directly to PLS-SEM.
Kurtosis outside an initial cutoff does not automatically mean that the data must be discarded. Record which variables you checked, the reason for each decision, the number of affected observations, and the results after processing. If you have no methodological reason to alter the data, retaining the valid data and reporting the limitation is usually more defensible than deleting observations in bulk.
Distinguishing Kurtosis from Skewness and Standard Deviation
Kurtosis and skewness often appear next to each other in the Descriptive Statistics table, so they are easy to confuse when you first run SPSS. Skewness measures the asymmetry of a distribution. If the tail extends toward the right, skewness is usually positive; if it extends toward the left, skewness is usually negative. Kurtosis focuses on peakedness, flatness, and the tails.
Standard deviation answers a different question: how far the values are from the mean. A variable with a large standard deviation does not necessarily have high kurtosis. Conversely, a variable with a moderate standard deviation can still have an outlier that increases kurtosis. You can also review the median when comparing the center of the data with the mean in a skewed distribution.
| Statistic | Main question | Quick reading in SPSS |
|---|---|---|
| Mean | What is the average value | Describes the center and is sensitive to extreme values |
| Std. Deviation | How dispersed are the data around the mean | Uses the same measurement unit as the original variable |
| Skewness | Is the distribution skewed left or right | A value close to 0 usually indicates less asymmetry |
| Kurtosis | Is the distribution peaked or flat, with heavy or light tails | A value close to 0 is usually close to normal under the excess kurtosis calculation |
| Median | What is the middle value after sorting the data | Useful when the distribution is skewed or contains outliers |
Do not confuse kurtosis with variance either. Variance is built from squared deviations from the mean, while kurtosis uses a higher power to describe the shape of the distribution. If you need the underlying formula, refer to the variance formula, then return to the SPSS table and read the correct statistic.
Common Errors
The first error is concluding that the data are normally distributed simply because kurtosis is close to 0. The distribution may still be skewed in one area, contain outliers, or have a shape that does not fit the assumption when viewed in a histogram. Treat kurtosis as a diagnostic signal, not as a certificate that the data meet every assumption.
The second error is using a cutoff without citing its source. If your thesis specifies an acceptable range for kurtosis, state the criterion, the variables to which it applies, and the calculation method. When no specific requirement exists, report the actual values and describe the additional checks instead of forcing the data into a preferred range.
The third error is deleting an item immediately because its kurtosis is high. Before removing the item, check coding, missing values, out-of-range values, boxplot, and histogram. If the data are valid, deleting an item because of one statistic can change the content of the scale.
The fourth error is confusing Kurtosis with Std. Error of Kurtosis, or looking only at whether the value is negative or positive without considering its absolute size. Read the column name, sample size, and related descriptive statistics together.
The final error is placing the kurtosis of an item in the section where you discuss regression residuals. Name the two objects correctly. When writing Chapter 4, state which variable distribution you checked and which analysis step the check supported.
Frequently asked questions
What is kurtosis in SPSS?
Kurtosis is a statistic that describes the peakedness, flatness, and tails of a data distribution. In SPSS, it usually appears in the Descriptive Statistics table beside Skewness, Mean, and Std. Deviation.
What level of kurtosis is good?
Kurtosis close to 0 usually indicates a distribution shape close to normal under the excess kurtosis calculation. However, there is no single number that fits every research project. Also inspect the histogram, boxplot, sample size, and requirements of your analysis.
Does negative kurtosis mean that the data are wrong?
Negative kurtosis indicates that the distribution tends to be flatter than normal. This does not by itself show that the data are wrong. Check the value range, missing data, coding, and charts before deciding how to proceed.
What are kurtosis and skewness used for?
The two statistics help you describe the shape of a distribution and detect signs that require further checking before you run an analysis. Skewness focuses on asymmetry, while kurtosis focuses on peakedness, flatness, and the tails.
Do I need to delete an item when kurtosis is outside the criterion?
Not immediately. First check data entry, coding, outliers, and the item content. Delete an item only when you have a clear methodological basis, and record the reason and effect of that decision in your analysis process.
Open your SPSS file again, run Analyze > Descriptive Statistics > Explore, save the Descriptive Statistics table, histogram, and boxplot for the variables you need to check, then note which variables have unusual kurtosis for the next analysis discussion. If you need to run and interpret the analysis using your own .sav or .csv file, M4 analysis by DoThesis supports this process.