What Is Secondary Data? How to Collect and Use It in Research
What is secondary data
Secondary data is data that an individual, organization, or agency collected earlier for a specific purpose and that you reuse for your own research question. Examples include company financial statements, government statistics, population data, industry reports, published research, transaction data, and internal records that you have permission to use.
The timing and original purpose of collection are the points you need to establish. When you create and distribute a questionnaire and receive responses from participants, you are working with primary data. When you download an existing dataset from a public source or receive an organization's archived data, you are working with secondary data.
Secondary data can be quantitative or qualitative. In quantitative research, you may work with numeric columns such as revenue, customer counts, growth rates, ratings, or monthly transaction volumes. You can process them in Excel, SPSS, R, or other software, depending on your research design. If you are just starting, you can read more about scientific research and what scientific research is.
A secondary-data workflow usually involves five tasks: identifying the variables you need to measure, finding a suitable source, checking quality, standardizing the data, and citing the source clearly in your thesis. Having data available does not mean that you can immediately place it in an .sav or .csv file and run an analysis. You still need to show that the data measures the intended concept and fits the scope of your study.
Why secondary data matters in quantitative research
Secondary data can answer questions for which distributing a new questionnaire would take too much time or would not be feasible. If your topic examines an industry's revenue trend over 10 years, annual statistical reports may be more suitable than asking consumers to remember all their past behavior.
In quantitative research, this type of data is commonly used in three ways. First, you can use it to describe the research context, such as market size, the number of companies, or the growth rate. Second, you can use secondary data as variables in regression, panel-data, or time-series analysis. Third, you can combine it with primary data to check or supplement findings from a questionnaire.
Before you run an analysis, answer four questions: who collected the data, when was it collected, does the variable definition match your topic, and is the unit of observation consistent? A report that says “sales growth” may not be equivalent to “actual revenue adjusted for inflation.” The concepts are related, but they may not be interchangeable.
The data structure also affects your method. Data organized by company and year may form panel data. Monthly data needs to be checked for continuity over time. Aggregated data by province only supports conclusions at the province level. It does not automatically support conclusions about individual people.
If you are developing your topic, place secondary data within the broader Scientific Research topic so that you can define the scope, unit of analysis, and resources you can access.
How much secondary data is sufficient
There is no single threshold that applies to every secondary-data project. “Sufficient” first means that the source fits the research question, the variable definitions are clear, the number of observations is adequate, and the origin of the data can be checked. Statistical thresholds should come only after you have identified the variables and data structure correctly.
| What to check | Reference point | Source |
|---|---|---|
| Observations per item in a scale | 5 to 10 observations per item | (Hair et al., 2010) |
| Regression sample size with m independent variables | Minimum 50 + 8m | (Tabachnick and Fidell, 2013) |
| Cronbach's Alpha | 0.7 or above; an exploratory scale may be considered from 0.6 | (Nunnally, 1978); (Hair et al., 2010) |
| Corrected Item-Total Correlation | 0.3 or above | (Nunnally and Bernstein, 1994) |
| KMO in EFA | 0.5 or above | (Kaiser, 1974) |
| Bartlett's Test in EFA | p-value below 0.05 | (Kaiser, 1974) |
| Total variance explained in EFA | 50% or above | (Hair et al., 2010) |
| Durbin-Watson in regression | Between 1 and 3 | (Field, 2013) |
If the data is a pre-calculated index, do not automatically run Cronbach's Alpha or EFA. These tests are designed for scales containing multiple items that measure the same concept. A column called “competitiveness index” that another agency calculated from several components is not the same as a Likert questionnaire that you designed yourself.
For a finite population and a sample-size calculation, you can consult the finite-population formula from (Yamane, 1967). The formula does not turn an unsuitable data source into suitable data. You still need to explain the quality of the definitions, scope, and way observations were selected.
How to read secondary data in the output
When you open a secondary-data file in SPSS, read both the variable description and the analysis output. In Variable View, check the variable name, variable label, data type, number of decimal places, and value labels. In Data View, check missing data, unusual values, and consistency across rows.
The example below uses an assumed dataset containing 120 monthly observations. This is illustrative output, not the result of an actual study.
| Descriptive Statistics | N | Minimum | Maximum | Mean | Std. Deviation |
|---|---|---|---|---|---|
| Doanh_thu_trieu | 120 | 82.00 | 245.00 | 146.38 | 34.71 |
| Chi_phi_marketing | 120 | 5.00 | 42.00 | 18.26 | 8.94 |
| Luot_khach | 120 | 1,240 | 8,950 | 4,816.25 | 1,732.40 |
| Valid N (listwise) | 120 |
In the Descriptive Statistics table, N shows the number of valid observations included in the calculation. If one variable has a much lower N than the others, check for missing data before running the regression. Minimum and Maximum help you detect incorrect codes, such as negative revenue when the variable definition allows only positive values.
If you run a regression through Analyze > Regression > Linear, the Model Summary table shows R Square and Adjusted R Square. The ANOVA table reports the overall model test, while the Coefficients table reports the coefficient for each variable. These tables are meaningful only when the data is arranged according to the correct unit of observation and the model fits the research question.
Create a separate data dictionary containing the variable name, definition, measurement unit, source, time period, and treatment of missing values. When your supervisor asks where a number came from, you can trace it from the output back to the data column and the original document.
What to do when secondary data is unsuitable
Work from source problems to technical problems. This order helps you avoid deleting large numbers of observations simply to make the output look better.
Recheck the variable definitions and scope
Read the original documentation and record the definition of each variable. Check the currency, time unit, geographic area, target group, and aggregation method. If the source variable does not represent the concept in your model, find a replacement variable or adjust the research question.
Check missing data and data codes
In SPSS, use Analyze > Descriptive Statistics > Frequencies or Analyze > Descriptive Statistics > Descriptives to inspect the basic distribution. Look for codes such as 99, 999, or -1, because they may mean “don't know” rather than a real value. Read the source documentation before converting these codes into missing values.
Standardize the format
Use one date format, standardize the measurement units, and give variables distinct names. If you combine data from two sources, create a crosswalk table before merging the files. Record every column-name change, because a new name in SPSS must still be traceable to the original name.
Handle outliers with a clear basis
A very large value is not automatically an error. It may represent a large company or a month affected by an unusual event. Remove an observation only when you have support from the source documentation, the collection rules, or a clear logic check. Record the row number, reason, and operation performed.
Find another source or collect new data
If the source does not report its method, omits important years, or does not allow its reliability to be checked, editing the file in SPSS will not solve the original problem. You can find another source or combine the secondary data with primary data. When you need to distinguish the two workflows, see the guide to primary data.
Distinguishing secondary data from primary data
The question “what are primary and secondary data” often appears when you write the methods chapter. The main distinction is who collected the data and what its original purpose was. The distinction does not depend on whether the data is numeric or textual. Primary data can be numeric Likert responses, while secondary data can also be found in written records.
| Criterion | Primary data | Secondary data |
|---|---|---|
| Collector | You or your research team | An agency, company, or other researcher |
| Original purpose | Designed for the current research question | Created for an earlier purpose |
| Examples | Online questionnaire, Likert scale | Industry report, statistical data, transaction records |
| Control | You can design the variables and collection procedure | Limited by the source's definitions and scope |
| Main risks | Respondent bias, low response rate | Limited methodological information, different variable definitions |
You can use primary and secondary data in the same study, but you need to separate their sources and collection times. Industry statistics can describe the context, while a questionnaire can measure respondents' perceptions. Do not combine the two types into one variable unless you have a basis for the measurement units and comparability.
You should also distinguish secondary data from qualitative methods. Secondary data describes the origin of the data, while qualitative research describes the form of the data and how it is analyzed. A published written report is secondary data, but how you use it depends on the research objective. You can read what qualitative research is to avoid mixing the two concepts.
Common mistakes
The first mistake is citing a general link without recording the report title, publication year, issuing organization, and page containing the data. Readers need to know exactly where a number came from, especially when the source has several updated versions.
The second mistake is treating every dataset downloaded from the Internet as reliable secondary data. An online summary may have lost its definitions or may have been copied through several sources. Prefer the original document and save the PDF, access date, file name, and data version.
The third mistake is changing the data by intuition to make the output meet a threshold. Deleting unusual observations, changing a missing code to zero, or replacing blank values with the mean can all change the result. Each operation needs a methodological reason and must be recorded in the data-processing log.
The fourth mistake is using aggregated data to make conclusions about individuals. Provincial revenue data supports analysis at the province level. It does not automatically support the conclusion that every person in the province behaves in the same way.
How to write this in your thesis
You can write: “The study uses secondary data from [name of agency or report], published in [year], comprising [number of observations] observations during [time period]. The variables [variable names] were standardized to [unit], missing data were handled using [treatment], and the origin of each variable is presented in [appendix or table].” Replace every bracketed placeholder with real information from your file.
Frequently asked questions
What is secondary data in scientific research?
It is data that existed before you conducted the current study and was collected by another individual or organization. Statistical reports, company data, industry reports, and published research can all be secondary sources when you reuse them for a new purpose.
Record the source, timing, scope, and way the data was created. A file name or general link alone is usually not enough to explain the data to your committee.
Where can I get secondary data?
You can look in government data portals, annual reports, industry reports, academic databases, authorized company records, or appendices to published research. Prioritize sources that clearly identify the issuing organization, time period, variable definitions, and collection method.
If you download data from an intermediary website, trace it back to the original source before putting it in your thesis. Save the version of the document that you used.
Does secondary data require Cronbach's Alpha?
Run Cronbach's Alpha only when your data contains multiple items designed to measure the same concept. An economic index or transaction dataset is not a Likert scale, so Alpha may not be appropriate.
If the source has published a scale and you are reusing its items, describe the scale structure clearly and check reliability according to your research design.
Should I combine primary and secondary data?
You can combine them when the two types of data address the same research question and have sufficiently compatible units, time periods, and definitions. For example, secondary data can describe the market, while primary data measures respondents' perceptions.
Present the source, processing procedure, and limitations of each type separately. Do not merge them into one dataset unless you have a valid way to link them.
What should I do if a lot of secondary data is missing?
First, check the missing codes and original documentation to distinguish blank cells from values of zero. Then assess the missing rate by variable, time period, and observation group. Possible responses include finding additional sources, removing a variable that lacks sufficient information, or using an appropriate method when your research design permits it.
Record the amount of missing data, the treatment criteria, and the effect on sample size. Do not replace every blank cell with zero simply so that you have enough data to run SPSS.
Open your data file and create a table containing the variable name, definition, measurement unit, source, time period, and missing code. Then compare each row with the original document. If you need to run the analysis on your own .sav or .csv file, you can use M4 Data Analysis.