SPSS Export Format

1. Overview

The SPSS format presents data in the same wide structure as the Wide format export (one row per respondent, one column per question) but also includes additional metadata for each variable. This makes it the preferred choice for researchers using SPSS, Stata, or other statistical tools that can read SPSS files.

Standard SPSS exports are delivered as .sav files, which open in SPSS as well as Stata, R, and other third-party tools that read SPSS files. The Q-optimized bundle uses compressed .zsav data with the same SPSS metadata.

2. File Structure & Layout

Standard wide SPSS data has one row per respondent and one variable per question, expanded option/topic or metadata field. SPSS variable names and wide Excel headers use the same stable reporting-ID naming rules.

The reporting ID identifies the source question. The exported column name is its normalized, file-safe identifier, with an indexed suffix when the question expands. The reporting label is the readable name used for analysis, while question text preserves the original wording. For example, how-old-are-you, how_old_are_you, Age and How old are you? are four different fields for the same question.

Names use letters, digits and underscores. Punctuation is normalized to underscores and outer underscores are removed. A name that does not start with a letter, or consists only of V followed by digits, receives a q_ prefix; an empty normalized name becomes column. Names are limited to 50 characters in these exports.

Expanded questions use _1, _2, and subsequent indexed suffixes for options or topic combinations. Numbering follows the survey-definition option and topic order, with deterministic natural ordering for values without a defined order. The suffix is a column index, not the respondent's selection rank. Changing readable labels does not rename the source ID; changing the question definition or option order can change its expansions.

Normalization, truncation and case-insensitive name collisions are resolved using deterministic hash suffixes derived from the source identifier. Indexed endings are retained where possible. Always use the supplied map or codebook to identify the resulting name rather than recreating it from a label.

Example (first 5 columns):

The standard export begins with respondent metadata such as respondent_id, status, weight, start_time, end_time and duration, followed by question variables. Available sequence and quality fields also belong to this leading metadata block. These variables are not prefixed with positional numbers; match them by name and codebook rather than fixed position.

When available, respondent_sequence is the respondent's survey-wide ordinal. Filtering respondents preserves the original sequence, so gaps are expected. A missing underlying marker can leave the value empty; older datasets without sequence metadata retain their previous variables. Look for the field in the accompanying codebook when present.

The following actual variable names match their codebook entries:

Exported column / SPSS variable	Stable reporting ID	Readable reporting label	Original question text	Option/topic
how_old_are_you	how-old-are-you	Age	How old are you?	—
quota_quads	quota-quads	Quads	quota-Quads	—
which_of_the_following_brands_have_you_heard_of_1	which-of-the-following-brands-have-you-heard-of	Which of the following brands have you heard of?	Which of the following brands have you heard of?	Diet Mountain Dew
which_of_the_following_brands_have_you_heard_of_2	which-of-the-following-brands-have-you-heard-of	Which of the following brands have you heard of?	Which of the following brands have you heard of?	Canada Dry Ginger Ale

In SPSS File and Codebook, map SPSS variable name to Reporting ID, Reporting label, Full question text and Option/topic suffix. Short SPSS label is the readable label carried in the SPSS file; it is not the stable variable name. Value/code labels explain coded responses. In this example, Age uses 1 = 55+, 2 = 35-54, and 3 = 18-34; do not interpret these category codes as ages.

Compatibility: these identifiers replace previous V###_descriptive_label headers. Update scripts, saved SPSS syntax, joins and BI imports that match the previous headers or readable labels. Use a fresh data map or codebook to build the old-to-new mapping, verify expanded options, and retain that mapping with the export. Do not assume an old positional variable number is a reporting ID.

3. Row Contents

The metadata fields identify the respondent, latest status, weight, start/end times and duration. Question fields hold codes or values appropriate to the selected export. Use the codebook for their types, labels and option mappings.

Respondent quality variables

Available quality_suspect_score, quality_duplicate_detected and quality_termination_reason variables use the latest recorded value per respondent and appear with metadata before survey-question variables. The suspect score retains its measured 0–5 value; it is not recoded into an answer-category index. Missing evidence stays missing, rather than becoming zero or a negative duplicate finding.

Use the codebook and value labels to interpret duplicate findings and termination reasons. The corresponding numeric-coded wide Excel export uses SPSS-compatible codes with a Data Map. See Respondent quality fields for why current evidence and recorded exclusion can differ.

4. Data Representation

Single-choice questions

Stored as one column per question with the selected answer recorded.

Multi-choice questions

Each option is represented as a separate column. The value assigned to each selected option indicates the order in which the respondent selected it (e.g., -1 = not selected, 1 = selected first, 2 = selected second, 3 = selected third, etc.). This preserves the sequence of selections, which can be useful for analyzing which options respondents considered first.

If the respondent selects an exclusive option (such as "None of the above" or "Prefer not to say") after previously selecting one or more non-exclusive options, the exclusive option is recorded with a value reflecting the order in which it was clicked, and all previously selected non-exclusive options are deselected at that point (reset to -1). Only the exclusive option will appear as selected in the final data.

You may want to consider defining MRSETS within SPSS to facilitate easy reporting, as this is not something that MX8 Labs includes in the export file.

Numeric questions

Stored directly as numeric values. "Don't know" responses are coded as -999 and flagged as an SPSS missing value, so they are automatically excluded from means, medians, and other numeric summaries.

Rating and NPS questions

Rating and Net Promoter Score (NPS) questions are stored as their numeric scale value (for example, a 1–5 rating or a 0–10 NPS score), so they can be analyzed directly as scale variables — means, top-box, and NPS calculations all work without re-coding. The descriptive value labels are retained alongside the numeric values, so labeled categories still display in SPSS and Stata.

Fractional scale values are preserved exactly as captured (for example, a 3.5 on a half-point rating scale is exported as 3.5, not rounded), so half-point and other non-integer scales analyze correctly.

Open-end questions

The standard coded SPSS export includes only the coded values for open-end questions, not the raw verbatim responses. This keeps the file aligned with a statistical analysis workflow where categorical codes are more useful than free text.

• Each coded open-end appears as a labeled numeric variable, with value labels mapping each code to its human-readable category.
• Until the open-end field has been closed (i.e., coding is finalized for the field), all responses in that variable are marked as "To be classified" in the export. Once the field is closed, the export reflects the final code assignments.
• If you need the raw, uncoded verbatim responses, download the data in CSV or one of the other non-SPSS formats, which include the full response text.

Use an uncoded SPSS export, Long Excel, or raw CSV when you need the original verbatim response instead of the finalized open-end code.

Uncoded exports

The SPSS format is also available as an uncoded export, which returns the underlying response value as it was captured rather than the recoded category. Use it when you want to analyze or re-derive the original answers, including open-ended verbatims, instead of the recoding applied for reporting. Select an Uncoded card in the download dialog; the variable structure and metadata are otherwise unchanged.

Q-optimized exports

Choose Q-optimized SPSS File and Codebook when you plan to import the data into Q and want multi-select questions detected as Pick Any questions.

The export creates two related groups for each multi-select question:

• Selection variables show whether each option was selected. Respondents who were not shown an option are excluded from that option's base.
• Selection-order variables retain whether an option was selected first, second, or later. These variables use _order in their names.

This preserves both the Pick Any base and the order in which options were selected, including for rotated or selectively exposed options.

The download is a ZIP archive containing a compressed .zsav dataset and an Excel codebook.

To check the import in Q:

1. Keep Automatically detect data file structure enabled when importing the .zsav file.
2. Confirm each multi-select selection group appears as one Pick Any question.
3. Check one respondent who saw all options, one who saw a subset, and one who was not shown the question.
4. Confirm the _order group retains first, second, and later selection positions.

Analysis-ready bundles

Choose SPSS Analysis Bundle when repeating responses need to be analyzed separately from one-row-per-respondent data.

After extracting the ZIP:

1. Start with bundle_index.xlsx. Its Datasets and Variables sheets explain what each file and variable represents.
2. Use respondents.sav for one-row-per-respondent analysis.
3. Use the relevant stack_*.sav file when analyzing a repeating response. Each row represents one respondent at one unique combination of the stack dimensions listed for that file.
4. Use the respondent details included in a stack file when you need demographics or weights in the same analysis.
5. To open every dataset in SPSS, set the working directory to the extracted folder and run open_all_tables.sps.

Use README.txt for guidance included with the bundle. Choose Uncoded SPSS Analysis Bundle when you need the original captured values instead of coded reporting values.

5. Metadata Provided

The SPSS format includes additional metadata that makes analysis easier:

• Variable labels - Readable reporting labels with option/topic context where applicable. Original question wording is retained separately in the codebook.
• Value labels - Mappings of codes to human-readable labels (e.g., 1=Male, 2=Female, 3=Non-Binary).
• Measurement levels - Nominal, ordinal, scale, etc., depending on the question type.
• Missing value definitions - Explicitly marked missing values (e.g., -1 = Not selected).
• Variable types - Numeric, string, date/time.

This metadata ensures the dataset is analysis-ready in SPSS and other statistical software.

6. Missing & Special Values

• -1 typically denotes unselected or non-applicable options.
• Empty cells may represent skipped questions.
• "Prefer not to say" appears as a standard category.
• All missing values are flagged as SPSS missing values in the export — including any negative sentinel codes (e.g., -1, -2, -99). This means SPSS, Stata, and other compatible tools will automatically exclude these values from calculations such as means, frequencies, and cross-tabs, without requiring you to define missing values manually.

7. Weighting

• Apply the weight column in analysis to ensure results reflect target population.

8. Best Practices

• Use the built-in metadata in SPSS/Stata to reduce manual labeling.
• Leverage variable labels to quickly identify questions.
• Use value labels to decode numeric response values.
• Use the codebook to match the exported variable to Reporting ID, then join to reporting_id in Long format. Include the option/topic mapping for expanded variables.

9. Stacked Exports

The SPSS format supports stacked exports, where the data is organized by the tags assigned to each question in the survey editor. In a stacked export, each respondent row is repeated for every tag group, and only the variables belonging to that tag are included alongside the respondent identifier and metadata variables (weight, timing). All SPSS metadata — variable labels, value labels, and measurement levels — is preserved for each variable in the stacked output.

To generate a stacked export, select the Stacked option in the download dialog. The resulting file will contain a Tag variable indicating which tag group each row belongs to.

This is particularly useful when you want to run separate analyses on different sections of a survey (e.g., "Brand Awareness" vs. "Purchase Intent") without manually subsetting the data.

10. Choose an SPSS export

Choose	When to use it	Rows in the file
SPSS File or SPSS File and Codebook	General work in SPSS, Stata, R, or another compatible tool.	One row per respondent.
Q-optimized SPSS File and Codebook	Q analysis that should detect multi-select options as Pick Any questions while retaining selection order.	One row per respondent, with separate binary selection and ordinal order groups.
Raw Data in SPSS (Stacked) or its codebook bundle	Analysis organized by survey tags where repeating respondents in one file is intentional.	A respondent can appear once per tag group.
SPSS Analysis Bundle	Complex repeating responses that should be separated into analysis files.	One respondent file plus separate files for each repeating structure.

Choose the corresponding Uncoded card when you need original captured values instead of the coded categories used in reporting. If you are unsure, start with SPSS File and Codebook; use Q-optimized for Q, or an analysis bundle when repeating responses need separate files.

For the download-dialog workflow, see Downloading data and reports.