What Is Communality in EFA?

You open the SPSS output after running EFA and see the Communalities table with the Initial and Extraction columns. The number on each item row shows how much of that item's variance is explained by the extracted factors. Read this indicator together with the Rotated Component Matrix, Total Variance Explained, and the theoretical quality of the scale. It should not be used on its own to decide whether to delete an item.

If you are reviewing the full procedure, see the guide to EFA and the resources in the Statistics topic. The discussion below focuses on the number currently appearing in your file.

What Is Communality

Communality, also called the communality value or the extracted variance of an item, is the proportion of an item's variance explained by the common factors after factor analysis. In SPSS, this indicator appears in the Communalities table.

For an item, communality is usually expressed as the sum of the squared factor loadings for that item on the retained factors:

h²ᵢ = λᵢ₁² + λᵢ₂² + ... + λᵢₖ²

Here, h²ᵢ is the communality of item i, and λ is the loading of that item on each extracted factor. If an item has loadings of 0,70 and 0,20 on two retained factors, its communality is approximately 0,70² + 0,20² = 0,53.

The value ranges from 0 to 1. A higher value means that less of the item's variance remains specific to that item and that the factors in the model explain the item more effectively. A low value suggests that the item may not fit the current factor structure, but it is not enough evidence to delete the item immediately.

In SPSS, Initial usually starts at 1 when you use Principal Component Analysis. With common factor extraction methods such as Principal Axis Factoring, the Initial value may be estimated differently. The column to focus on when evaluating the post-extraction result is Extraction.

Why Communality Matters in Quantitative Research

Communality answers a practical question: how much of each item's variance is explained by the factors that were retained. If you have a service-quality scale with several items, communality indicates whether the extracted factor structure preserves the content of each item.

This indicator helps you identify three situations. First, an item may have low loadings on every factor. Its communality is usually low because no factor explains it well. Second, an item may load strongly on several factors, creating cross-loading. Its communality can still be high, so looking at communality alone would miss a discriminant validity problem. Third, an item may have an acceptable communality but still fail to fit the theoretical meaning of the factor.

Read communality in the following sequence: check KMO and Bartlett, identify the number of retained factors, read Total Variance Explained, review the rotated matrix, and then compare the communality of each item. A KMO of 0,5 or above and a statistically significant Bartlett's test below 0,05 are commonly used to assess whether the data are suitable for EFA (Kaiser, 1974). These are conditions for the correlation matrix as a whole, whereas communality provides item-level information.

Communality is also different from total variance explained. Total variance explained tells you how much variance in the complete set of items is explained by the factors. Communality tells you how well the factors explain each individual item. A model can have good total variance explained and still contain one or more items with low communality.

What Communality Value Is Acceptable

In many theses, a communality of 0,5 or above is treated as a useful reference level. You should present this as a practical evaluation criterion rather than an automatic rule. The sources in the allowlist specify factor loadings from 0,5 and total variance explained from 50%, while the decision about each item should also consider its loading, cross-loading, and theoretical basis (Hair et al., 2010).

The table below separates indicators that are often confused under the label “extracted variance.” Each row includes a source so you can explain the basis if your supervisor asks.

How you read this table matters. A communality of 0,48 does not automatically mean that the item must be deleted. If the item has clear content meaning, an acceptable factor loading, no serious cross-loading, and deleting it would remove important scale content, you can retain it and report the limitation. By contrast, a communality of 0,25 combined with a low loading is a stronger reason for review.

When writing your results, do not simply state that “all communalities were acceptable” without reporting the lowest value and explaining how borderline items were handled. Report the actual values from your output, not the illustrative values in this article.

How to Read Communality in the Output

In SPSS, go to Analyze > Dimension Reduction > Factor. Move the items into the Variables box, choose the extraction method in Extraction, select the rotation method in Rotation, and run the EFA. If you use Principal Component Analysis, the results usually display a Communalities table with the Initial and Extraction columns.

Reading the Initial Column

Initial is the starting value before the factors are extracted. With PCA, you will usually see 1,000 for every item because PCA treats all of an item's variance as variance being analyzed. Do not use this column alone to conclude that an item is well or poorly explained.

If you switch to a common factor extraction method, the Initial value may not equal 1. In that case, the difference between Initial and Extraction reflects the portion of variance retained by the model after only a limited number of factors have been extracted.

Reading the Extraction Column

Extraction is the main column for evaluating communality after EFA. For example, if the GC3 row has Extraction = 0,642, the retained factors explain approximately 64,2% of the variance of item GC3 in the illustrative dataset. The remainder is item-specific variance and error that the model does not explain.

The table below is illustrative output, presented in the format of SPSS output. It is not the result of a real study, and you must not copy these numbers into your thesis.

In this example, DV3 has a communality of 0,471, below the commonly referenced level of 0,5. You should not delete DV3 because of one number. Open the Rotated Component Matrix and check DV3's loading. See whether the item loads on the expected factor or loads on several factors. Then review whether respondents may have understood the item differently from what you intended.

If you use PCA, distinguish it from EFA according to your research objective. PCA finds components formed from total variance, whereas EFA focuses on exploring a latent factor structure. You can read more about Principal Component Analysis when your output uses this method.

Comparing Eigenvalue and the Rotated Matrix

The number of retained factors directly affects communality. Retaining more factors usually changes the total proportion of variance explained for each item. Therefore, if you change the factor-retention criterion or the extraction method, run the analysis again and read the complete new output. Do not combine the Communalities column from an earlier run with the rotated matrix from a later run.

You can also review how to read Eigenvalue. An eigenvalue greater than 1 is a reference rule for retaining a factor (Kaiser, 1960), but you still need to consider theoretical meaning and the loading structure. Communality does not replace the decision about the number of factors.

What to Do When Communality Is Low

Work from the least intervention to the most intervention. The goal is to address the actual cause and preserve the analysis history for explanation, rather than deleting items repeatedly until the table looks “good.”

Check the Extraction Method and Number of Factors

Confirm that you are using the method specified in your research design. Check the number of retained factors, the rotation method, and whether the item list was entered correctly. A change in Extraction or Rotation can change communality.

If you are using PCA only because SPSS set it as the default, discuss with your supervisor whether it fits your objective of exploring latent factors. Do not change the method just to raise one indicator.

Check Loading and Cross-Loading

Open the Rotated Component Matrix. If an item has a low loading on the expected factor or has two similar loadings on two factors, the problem may lie in the scale structure rather than in communality alone.

In some output, you need to enable the option to sort loadings by size and hide small coefficients so that the table is easier to read. Save the table before deleting an item so you have evidence for each analysis round.

Check the Content and Original Data

Review the coding, reverse-coded items, and missing values. A reverse-coded item that has not been recoded can sharply reduce its correlation with the other items in the same scale. Also check whether you entered the wrong column, used an incorrect value code, or retained invalid response rows.

If communality is low but the item is theoretically essential, you can retain it and state the reason clearly. If the item is low, has a weak loading, and reduces scale reliability, deleting it is usually easier to explain. Then rerun EFA and Cronbach's Alpha with the remaining items, and record which item was removed in which round.

Do Not Delete Items in Bulk

Deleting several items at once makes it difficult to know which change produced the new result. A safer approach is to review one item at a time, save the output, record communality, loading, cross-loading, and the reason for the decision. The committee usually cares more about whether the procedure is justified than whether the final table is perfect.

Distinguishing Communality from Similar Indicators

Communality is different from factor loading. Loading is the relationship between an item and one factor. Communality is the total portion of the item's variance explained by all retained factors, so an item with several component loadings can have a higher communality than the square of its largest loading.

Communality is also different from Total Variance Explained. Communality belongs to an individual item. Total Variance Explained belongs to the factor model as a whole. If total variance explained reaches 50% or above under the commonly used criterion (Hair et al., 2010), that still does not prove that every item has good communality.

Communality differs from eigenvalue. Eigenvalue represents the amount of variance explained by one factor across the full set of items. Communality represents the amount of variance explained by the factors for one specific item. You can read the eigenvalue article if you are confusing these two columns.

Communality is also not Cronbach's Alpha. Alpha assesses the internal consistency of a scale, whereas communality assesses how well the factor structure explains an item. A scale can have a high Alpha and still have an unsuitable factor structure, and the reverse can also occur.

Finally, communality in EFA is not a CFA indicator. When you move to CFA, you usually examine standardized loading, Composite Reliability, AVE, and discriminant validity. If your study includes a measurement-model confirmation stage, read the CFA guide instead of using EFA communality as a substitute for CFA evidence.

Common Mistakes

Looking Only at Communality to Delete an Item

This is the most common mistake. A low value must be read together with loading, cross-loading, scale content, sample size, and the research objective. Deleting an item according to a mechanical threshold can remove an important aspect of the construct.

Using Initial to Draw a Conclusion

With PCA, the Initial column is usually 1,000, so it does not help distinguish which items are well explained. Focus on Extraction after confirming the extraction method and number of factors.

Confusing Communality with Total Variance Explained

One indicator belongs to an individual item, while the other belongs to the full set of items. When writing Chapter 4, report them in separate passages and use the exact SPSS table names.

Copying a Threshold Without Giving the Source

If you use 0,5 as the loading or communality reference in your explanation, cite the appropriate source (Hair et al., 2010). Do not place several thresholds from different sources in one paragraph without explaining which indicator each threshold applies to.

Failing to Save Each Run

After every item deletion, save the output file or export the results table. You need to show which item was deleted, why it was deleted, and how the factor structure changed.

Frequently asked questions

What communality value is acceptable in EFA?

A value of 0,5 or above is commonly used as a reference when assessing the portion of an item's variance explained by the factors. It is not the only basis for a decision. You also need to examine factor loading, cross-loading, content meaning, and scale reliability, and cite the source (Hair et al., 2010) when using related thresholds.

Does low communality mean that I must delete the item?

Not necessarily. If the item has a clear theoretical role and the other indicators are acceptable, you can retain it, explain the decision, and report the limitation. If low communality occurs together with low loading or cross-loading, considering deletion has a stronger basis.

Why are all my communality values 1,000?

You may be looking at the Initial column in the Communalities table while using PCA. Check the Extraction column because it contains the value after the factors have been retained. If the entire Extraction column is also 1,000, review the output and the extraction method.

Is communality the same as extracted variance?

The two concepts are related but are not reported in the same way. Communality is the variance explained for each item. Total Variance Explained is the proportion of variance explained for the complete set of items by the retained factors.

Can I use communality to prove that a scale is acceptable?

Communality is only one part of EFA assessment. To evaluate the scale, you also need to examine KMO, Bartlett, the number of factors, loading, cross-loading, total variance explained, and consistency with the theoretical basis. Do not write that the scale is acceptable only because all items have high communality.

Open the Communalities table in your SPSS file, record the Extraction value for each item, and compare it with the Rotated Component Matrix before deciding whether to delete or retain an item. If you need to run EFA on your own .sav or .csv file and turn the results into thesis writing, M4 data analysis is the module for this step.