What Is Skewness? How to Read and Interpret It in SPSS

What Is Skewness

You open SPSS, check the Descriptive Statistics table, and see a Skewness column for each variable. This statistic tells you whether the distribution of observed scores leans to the left or right relative to a symmetrical shape. In practical terms, skewness helps you see whether most responses are concentrated toward one side of the scale.

For a random variable, skewness is commonly described using the standardized third central moment:

Skewness = μ3 / σ³

Here, μ3 is the third central moment and σ is the standard deviation. You do not need to calculate this formula yourself when you already have a .sav file. SPSS calculates it from the valid values for the variable. If skewness equals 0, the distribution is theoretically symmetrical. A positive value means the distribution has a longer tail toward the right, while a negative value means the tail extends toward the left.

You should read skewness together with kurtosis, a histogram, a Q-Q plot, and the measurement context. One number by itself is not enough to decide whether your data have a problem. You can also review Statistics topics if you are checking descriptive statistics before running regression, EFA, or PLS-SEM.

Why Skewness Matters in Quantitative Research

In quantitative research, skewness mainly helps you assess the shape of the distribution for an observed variable or a composite variable. This information can affect how you check assumptions, choose a statistical test, and interpret means. For example, if almost every respondent selects 5, the variable may be negatively skewed because the scores are concentrated at the high end of the Likert scale.

Positive skewness often appears when many observations are in the low-score range but a small number of observations have very high scores. Negative skewness often appears when many observations are in the high-score range and a small number have very low scores. The sign of skewness shows the direction of the skew, while the absolute size shows whether the skew is mild or strong.

You should distinguish skewness from standard deviation. Standard deviation describes how dispersed the observations are around the mean, while skewness describes asymmetry in the distribution's shape. A variable with a large standard deviation does not necessarily have large skewness, and a variable with skewness close to 0 may still be widely dispersed.

Skewness also supports your data-checking step before analysis. If the value is highly skewed, review the variable coding, unusual values, reverse-coded items, and the possibility of a ceiling or floor effect. In many social science studies using Likert scales, however, moderate skewness does not automatically mean that you must delete an item or transform the data.

What Skewness Value Is Acceptable

The question of what skewness value is acceptable has no single threshold that applies to every research design. The appropriate threshold also depends on sample size, the purpose of the analysis, the type of variable, and how you assess the distribution. When you report the result, state the rule you selected rather than claiming that the data pass simply because skewness falls within a familiar range.

Some sources use the absolute value of skewness to describe the degree of asymmetry. The table below summarizes common interpretations in practice and gives the source for each rule. Treat these as reference points for reading output, not as permission to ignore charts or other assumptions.

These reference points do not replace a specific test. In linear regression, you must also consider residuals, linearity, constant variance, and outliers. In EFA, KMO, Bartlett's test, factor loadings, and total variance explained each have their own role. In SmartPLS, you should not include CB-SEM fit indices such as CFI, TLI, or RMSEA in the PLS-SEM assessment.

Sample size also changes how you view skewness. With a large sample, a small deviation can produce a very small p-value in a distribution test even when the chart does not look especially unusual. With a small sample, the test may have limited power to detect skewness. Combine the numerical result, the chart, and the purpose of your model.

How to Read Skewness in SPSS Output

To view skewness in SPSS, open the data file and select Analyze > Descriptive Statistics > Descriptives. Move the variables you want to check into Variable(s), click Options, select Mean, Std. deviation, Minimum, Maximum, Skewness, and Kurtosis, then click Continue and OK.

To view additional charts, select Analyze > Descriptive Statistics > Explore. Move the variable into Dependent List, click Plots, and select Histogram and Normality plots with tests. SPSS will produce a Descriptives table, a histogram, and a Q-Q plot. As you read the output, check whether the number of valid observations has fallen because of missing values.

The table below is illustrative output, not the result of a real study. The column names follow the Descriptive Statistics table that SPSS commonly produces.

In the CL1 row, skewness is -0.48. The negative sign means that the distribution is slightly skewed toward the higher scores, while the absolute value of 0.48 indicates a limited degree of skew under the reference rule above. The CL2 row has positive skewness of 0.36, so the direction is the opposite, but the skew is also mild.

The HL1 row has skewness of 1.27. You should not delete the variable immediately. Open the histogram to see whether a group of low scores creates a right-hand tail, check the frequency table for coding errors, and consider whether this item is difficult enough that respondents tend to choose low scores. Kurtosis of 1.86 also suggests that you should inspect the distribution, but kurtosis is different from skewness and must be interpreted separately.

Std. Error of Skewness shows the standard error of the skewness estimate. One additional check is to calculate z-skewness as Skewness / Std. Error of Skewness. You should not, however, use a mechanical division as your only basis for a decision. With survey data, compare the result with the histogram, Q-Q plot, and the theory behind the scale.

What to Do When Skewness Is Too High

First, confirm that the value is real. Check Variable View to see whether Type, Decimals, Values, and Missing have been defined correctly. An entered code of 99 that has not been declared missing can substantially distort the mean, skewness, and kurtosis. If the variable is a reverse-coded item, check whether you recoded it before calculating the composite score.

Check the Original Data and Unusual Values

Run Analyze > Descriptive Statistics > Frequencies to review each response level and the number of observations. Then check extreme cases with a boxplot under Analyze > Descriptive Statistics > Explore. You can read more about outliers to distinguish a valid but rare value from a data-entry error.

If you find an entry error, correct it using the questionnaire or original data and record the change. If it is a valid response, do not delete it simply because it makes the skewness look better. When you remove a case, you need a methodological reason and must preserve the earlier version of the data file.

Check Coding and Composite Variables

An observed variable may be strongly skewed while the average of several observed variables has a more stable distribution. You can calculate a composite variable after checking scale reliability and structure, but do not combine variables simply to pull skewness closer to 0. The combination must fit your conceptual model and the scoring procedure defined in advance.

If the questionnaire design creates a ceiling or floor effect, describe the phenomenon in your thesis. A log transformation, square-root transformation, or winsorizing can change the interpretation and is not always suitable for Likert variables. If you use a transformation, state the formula and the reason, then check the results again after the transformation.

Consider an Alternative Analysis Approach

When your goal is to compare means or run regression, assess the degree of skew together with sample size and the model residuals. Slightly high skewness does not automatically invalidate the entire analysis. If the analysis is sensitive to the distribution or there are serious outliers, discuss an appropriate option with your supervisor instead of deleting many rows yourself.

For PLS-SEM, the focus is on the measurement model and structural model, including outer loading, CR, AVE, HTMT, VIF, path coefficient, and bootstrapping. Skewness is not the only criterion that determines whether a PLS-SEM model is acceptable. Report the correct set of statistics for the method you selected.

Distinguishing Skewness from Kurtosis and Standard Deviation

Skewness and kurtosis both describe the shape of a distribution, but they answer different questions. Skewness focuses on asymmetry and the direction in which the tail extends. Kurtosis relates to peakedness or tail heaviness compared with a reference distribution. Therefore, you cannot use kurtosis in place of skewness when interpreting the output.

Standard deviation measures how far observations are dispersed around the mean. You can review variance to understand the relationship between variance and standard deviation, or see the variance formula if you need to present the method. Neither statistic tells you whether a distribution is skewed to the left or right.

The mean, median, and mode also help you identify the shape of a distribution. In a symmetrical distribution, these measures are often close to one another. In a skewed distribution, the mean may be pulled toward the longer tail and away from the median. You can read more about the median, but you should still inspect the histogram instead of reasoning only from three summary values.

A common error is to see skewness of -0.8 and conclude that the variable has poor quality because the sign is negative. The negative sign only indicates the direction of the skew, not whether the variable is good or bad. Another error is confusing Std. Error of Skewness with skewness. When reporting the result, record the value in the Skewness column and use the standard error only when you have a clear analytical reason.

Common Errors

Looking at one threshold and declaring the distribution normal. Skewness close to 0 does not prove that the entire distribution is normal. Check kurtosis, the histogram, the Q-Q plot, and the purpose of the analysis as well.

Deleting an item or row to make the statistic look better. An extreme observation may be real data. Remove it only when you have a defined criterion, evidence of a data-entry error, or clear evidence that the response is unusable.

Confusing the direction of the skew. Positive skewness means that the tail extends toward higher values, while most observations may still be in the low range. Do not simply describe most scores as high because you see a positive sign.

Using a skewness threshold for kurtosis. The two statistics have different meanings. If you report both in Chapter 4, state the column name and interpret each statistic separately.

Failing to check missing-value codes. Codes such as 99, 999, or 0 may have been entered to represent no response. If they have not been declared missing, the output will include those codes.

Reporting values that do not match the file. After each data edit, save a new version with a clear name and record N, the variable edited, and the reason. This helps you explain why output tables in different chapters do not match.

Frequently asked questions

What is skewness in SPSS?

Skewness in SPSS is a descriptive statistic showing the degree of asymmetry in a variable's distribution. In the Descriptive Statistics table, read the value in the Skewness column, not the Std. Error of Skewness column.

What skewness value is acceptable?

There is no single benchmark for every study. You can use sourced reference ranges, such as an absolute value below 1, which is often considered mild to moderate skew (Field, 2013), and then check the chart and the purpose of the analysis.

Does negative skewness mean that the data are wrong?

Negative skewness only means that the distribution's tail extends toward lower values. If the degree of skew fits the response context and there is no coding error, the data are not automatically wrong. Check the histogram and frequencies before deciding how to handle them.

What are skewness and kurtosis used for?

Both describe the shape of a distribution, but skewness concerns asymmetry while kurtosis concerns peakedness and the distribution tails. They support data checking, but they do not replace the full assumption checks for regression, EFA, or other models.

Does high skewness mean that I need to delete the variable?

Check data-entry errors, missing values, reverse-coded items, outliers, and ceiling or floor effects first. If the value is valid, consider the purpose of the analysis and discuss the option with your supervisor. Deleting a variable simply to place skewness within a preferred range gives you a weak explanation.

How should I write skewness in my thesis?

You can adapt the following sentence to your file: “Descriptive statistics showed that [variable name] had skewness of [value], indicating a [left-skewed/right-skewed/approximately symmetrical] distribution. This result was assessed together with the histogram and Q-Q plot before [analysis] was performed.”

Open your SPSS file now, run Analyze > Descriptive Statistics > Explore, compare the Skewness column with the histogram, and record the variables that need checking before your next analysis. If you want to run the analysis on your .sav or .csv file and turn the results into thesis-ready writing, see DoThesis M4 data analysis.