How can I save a Stata dataset so that it can be read by a previous version of Stata?

The format sometimes changes between releases to allow for new features. This format change can cause problems when two colleagues are working on a project and one of them has not upgraded to the latest release. For example, when a Stata 10 user tries to open a Stata 12 dataset from a colleague, they are presented with an error,

        . use auto.dta
        file auto.dta from a more recent version of Stata
        r(610);

There is, however, an easy solution to this problem. First, Stata 11 can read Stata 12 datasets, and Stata 10 can read Stata 11 datasets. Also, since Stata 4, StataCorp has provided a way to save your dataset so it can be read by the previous version of Stata. That is, you can save Stata 12 datasets so they can be read by Stata 8 through Stata 10. Stata 11 or Stata 10 datasets can be saved so they can be read by Stata 8 or Stata 9 (Stata 8 and Stata 9 share the same dataset format). Stata 8 or Stata 9 datasets can be saved so they can be read by Stata 7. Stata 7 datasets can be saved so they can be read by Stata 6, and so on down to Stata 4 datasets, which can be saved so that Stata 3.1 can read them.

Saving datasets in Stata 12 so they can be read by Stata 11

Stata 11 can read Stata 12 datasets, so you do not need to do anything special:

        . save auto
        file auto.dta saved

Saving datasets in Stata 12 so they can be read by Stata 8, Stata 9, or Stata 10

To save a dataset in Stata 12 so that it can be used in any of Stata 8 through Stata 10, use the saveold command.

        . saveold autoold
        file autoold.dta saved

Saving datasets in Stata 10 or Stata 11 so they can be read by Stata 8 or Stata 9

To save a dataset in Stata 10 or Stata 11 so that it can be used in Stata 8 or Stata 9, use the saveold command.

        . saveold autoold
        file autoold.dta saved

Saving datasets in Stata 9 so they can be read by Stata 8

Stata 8 and Stata 9 datasets have the same format. Therefore, a dataset saved in Stata 9 can be used in Stata 8 with no problems or extra work.

The only issue related to this is that Stata 9 allows value labels to be up to 32,000 characters long. If Stata 8 tries to read a Stata 9 dataset with value labels that exceed the Stata 8 limit (244 for Stata/SE; 80 for Stata Intercooled), Stata 8 will ignore those labels and read the rest of the dataset.

Saving datasets in Stata 8 or Stata 9 so they can be read by Stata 7

To save a dataset in Stata 8 or Stata 9 so that it can be used in Stata 7, use the saveold command.

        . saveold auto7
        file auto7.dta saved

Saving datasets in Stata 4 to Stata 7 so they can be read by the previous version

To save a dataset in Stata 4, 5, 6, or 7 so that it can be read by the previous version, use the save command with the old option.

        . save auto6, old
        file auto6.dta saved

The old option in Stata 7 will not automatically truncate long variables or label names. This must be dealt with in one of two ways. You can either rename the variables and labels by hand or use Stat/Transfer to translate the datasets. The second suggestion is very handy if you have many variables.

Also, if any of your variables have value labels and the name of the value label is longer than 8 characters, you must either drop or rename those value labels.

        . label list
        origin:
                  0 Domestic
                  1 Foreign

        . label save origin using mylabel.do
        file mylabel.do saved

This command will produce a do-file called mylabel.do with the following commands:

        -------------------------------------------------mylabel.do
        label define origin 0 `"Domestic"', modify
        label define origin 1 `"Foreign"', modify
        -------------------------------------------------mylabel.do