How to Get Data from the General Statistics Office for a Quantitative Thesis

What does getting data from the General Statistics Office mean

Getting data from the General Statistics Office is the process of finding, downloading, checking, and citing statistical data published by the statistical authority. In a quantitative thesis, this is usually secondary data, used to describe the research context, construct province-level indicators, or provide inputs for analysis across time, regions, and population groups.

You need to distinguish between two types of data from the beginning. Aggregate data, such as population, retail sales, or the number of active enterprises, usually appears in publicly available statistical tables. Microdata, such as individual household responses, is often not freely downloadable as a complete personal-level file. Therefore, do not write in your proposal that you will obtain individual-level data from the General Statistics Office until you have confirmed that the dataset is accessible.

In practice, the process has five parts: identify the variables you need, find the table with the correct indicator, download or copy the data, check its consistency, and record the source and access date. If the final file will be analyzed in SPSS, standardize the variable names, years, measurement units, and province codes before opening the software.

Why General Statistics Office data matters in quantitative research

Data from the General Statistics Office helps you answer questions that an individual questionnaire cannot answer well. For example, you can describe market size using total retail sales, compare growth rates between localities, or include macroeconomic indicators in a regression model. This data also gives the “research context” section concrete evidence instead of general statements.

However, province-level data does not automatically become individual-level data. If your dependent variable is each person’s purchase intention, a province-level population table only describes the context. It does not replace the indicators in your questionnaire. You need to state clearly whether the unit of analysis is an individual, household, enterprise, province, or year.

Before searching for data, draw a small table with three columns: variable name, unit of analysis, and expected source. This prevents you from taking a province-level indicator and directly merging it into a file where each row represents one respondent without explaining how the merge works.

Secondary data can also help you check whether your primary data is plausible. If the questionnaire shows a very high service-use rate but the local context has low access, check the sampling method, screening questions, and data-entry process again. This check does not prove that a dataset is right or wrong, but it helps you identify points that need to be explained.

If your file contains values that are very different from the rest, see the guide to outliers in quantitative data. If you are building measurement variables from a questionnaire, the article what is a scale will help you separate the construct, indicators, and measurement unit before combining the secondary data.

How much General Statistics Office data is sufficient

For General Statistics Office data, “sufficient” does not simply mean having many rows. A suitable table must match the concept, use the correct unit, cover an appropriate time series or geographical scope, and provide a source that readers can check. You also need to see whether the definition of the data changed across years.

The thresholds below apply to quantitative decisions that commonly arise when preparing thesis data. They are not standards for evaluating the quality of the General Statistics Office. Each row states the source so you can explain the basis for the decision in the methods chapter.

If you only use aggregate data by year, do not mechanically apply questionnaire sample-size thresholds. A file containing 63 provinces over 5 years produces 315 rows technically, but those data still have group and time structures. In that situation, explain how you selected the unit of observation and consider an appropriate method instead of calling 315 rows 315 respondents.

In addition to quantity, check four criteria: whether the indicator matches the research concept, whether the measurement unit is consistent, whether the time period is continuous, and whether the table notes any change in calculation method. A table for the “number of enterprises” and a table for “active enterprises” may cover different populations even though their names sound similar.

How to read General Statistics Office data in output

General Statistics Office data usually does not have one output table like SPSS. After you enter the data, SPSS produces descriptive and test tables based on the file you standardized. Therefore, before reading Descriptive Statistics, you must know what each column represents and what unit of observation each row represents.

The example below assumes that you entered retail sales data for three provinces across three years. The numbers are illustrative only. They are not the results of a real study and do not represent data published by the General Statistics Office.

illustrative output

In SPSS, each row above can represent one province-year observation. Go to Analyze > Descriptive Statistics > Descriptives to view N, Minimum, Maximum, Mean, and Std. Deviation for the numeric variables. If Ban_le_ty_dong has an N lower than the number of rows you entered, check for blank values, the “...” symbol, or cells containing text.

Next, go to Analyze > Descriptive Statistics > Frequencies if you want to check the number of provinces and years. With panel data, check whether a province was duplicated incorrectly, a year was entered as a string, or two different names were used for the same locality. For example, “TP HCM” and “Ho Chi Minh” may be treated as two groups by SPSS if you entered them as strings.

If you run a regression, the path is Analyze > Regression > Linear. Put the dependent variable in Dependent and the explanatory variables in Independent(s), then open Statistics and select Estimates, Model fit, Collinearity diagnostics, and Durbin-Watson. The Coefficients table gives the coefficient for each variable, while Model Summary gives R, R Square, and Adjusted R Square.

You should save three separate files: the original downloaded file, the cleaned file, and the .sav file with variable labels. If someone asks why your mean differs from the published table, you can return to the original file and check the unit conversion, rounding method, or geographical scope.

What to do when the data is not suitable

First, identify whether “not suitable” refers to the source data or to the analysis result. If a table is missing one year, the problem concerns data scope. If SPSS reports many missing values, the problem concerns data entry and coding. If Cronbach's Alpha is low, the problem concerns the scale and cannot be solved simply by downloading another statistical table.

Check the indicator name and table notes again

Open the table title, measurement unit, geographical scope, and notes. Read the symbols below the table as well, because a dot, ellipsis, or the word “estimate” may have a different meaning. If the indicator does not match the variable definition in your research, remove that table from the model instead of renaming the variable to make it fit.

Standardize units and formats

Convert all monetary values to the same unit, such as million dong or billion dong. Remove thousands separators if SPSS is reading the number as a string, but keep an original copy for comparison. For percentages, determine whether the data is recorded as 4.2 or 0.042 before using it in a formula.

Record missing data instead of guessing

Do not turn a blank cell into 0 unless the source states that the value is zero. You can code missing values as system-missing in SPSS and report how many observations were excluded from each analysis. If missing data is concentrated in one province or period, state this limitation clearly in the thesis.

Find an alternative table or narrow the question

If the indicator you want is unavailable, find the closest alternative, but explain the difference. When the time series is not long enough, you can narrow the year range or change the objective from trend analysis to a description of the current situation. A narrower, explained scope is usually safer than combining sources with different definitions.

You should also avoid deleting provinces with high values simply because they change the mean. Check whether the value is an entry error, a real difference in scale, or a valid outlier. If you are analyzing mediation or moderation, distinguish clearly between the original variable, mediator, and interaction variable. These concepts are illustrated further in the articles about mediator and moderator.

Distinguishing General Statistics Office data from primary data

General Statistics Office data is secondary data, collected and published by an organization before you reuse it. Primary data is collected directly from the research subjects through a questionnaire. Both types can appear in one thesis, but you should not treat them as the same type in the methods section.

Another common confusion is treating the “total number of observations” in a statistical table as the survey sample size. Thirty province-year rows do not have the same meaning as 30 respondents. If the model requires individual-level data, you need a questionnaire and a separate collection process.

Similarly, descriptive statistics are not a scale. A percentage for a province cannot automatically become an indicator measuring an individual’s perception. You can use it as a context or control variable if the model and unit of analysis allow this.

Common mistakes

The first mistake is searching with a broad keyword such as “Vietnam revenue” and taking the first table that appears to fit. Start with the indicator name, unit, year, and geographical scope. Once you find a table, read the full title before downloading it.

The second mistake is saving only a screenshot. A screenshot may help with illustration, but it is inconvenient for analysis and often lacks information about the table version. Save the downloaded file or copy the data into a spreadsheet, and record the table name and access date at the same time.

The third mistake is using variable names that are too short, such as X1 and X2, and then forgetting their original meaning. In SPSS, use short names but add the full indicator, unit, and year in Variable Label. An easy-to-check naming rule is banle_2022_ty, dan_so_2022_nghin.

The fourth mistake is merging province data into an individual-level file without a merge key. If you need to merge the files, use a shared variable such as province code and year, check the number of rows before and after the merge, and confirm that no duplicate records were created. Save a copy before using Merge Files in SPSS.

The fifth mistake is citing “General Statistics Office” without naming the table. That description is too broad for readers to trace the source. In Chapter 3 or in the table note, state the agency, dataset or table name, data year, measurement unit, and access date.

Frequently asked questions

Where can I get data from the General Statistics Office?

Start with the official data portal or statistical publication website of the General Statistics Office, then search by indicator name, year, and geographical scope. When you open a table, read the title and notes before downloading it.

If a search result leads to a summary article, trace it back to the original table. The summary may have rounded the figures or quoted only part of the time series.

Can General Statistics Office data be used in SPSS?

Yes, if the data is converted into a structure that SPSS can read, such as one row for each province-year and one column for each variable. You need to process blank cells, text symbols, thousands separators, and data types before running the analysis.

If each row represents a respondent but you only have province-level data, keep the province data as a context variable and explain how it was merged. Do not treat it as individual responses.

Do I need to cite General Statistics Office data?

Yes. State the agency name, table or dataset name, year, measurement unit, data scope, link if available, and access date. The source should appear directly below the table in the results chapter and should also be described in the methods section.

What should I do if the data is missing?

First, check the notes to determine whether a blank cell means data is unavailable, not applicable, or zero. Then code the status correctly in SPSS and report the number of missing observations.

If too much data is missing in one period, you can narrow the scope or choose an alternative indicator with a closer definition. Do not enter 0 simply to fill every row.

Can General Statistics Office data replace a survey?

Only when your research question requires aggregate data at that same unit of analysis. If the topic measures individual perceptions, attitudes, intentions, or behavior, you will usually still need primary questionnaire data.

Review the model before deciding. If it includes a mediator or moderator, replacing individual data with province-level data can completely change the meaning of the model.

How to write this in your thesis

Open your current data file, make a list of the variables you need, and record the table name, unit, year, and citation method for each variable before entering it into SPSS. If you need to run the workflow on your own .sav or .csv file, you can use M4 data analysis from DoThesis.