Yes but when you figure out the meaning of each estimate that meaning will be the same.David Booth

Yes, but if you are categorizing your variables based on known categories mainly recorded in published researchs, books and national data record forms, It is good to categorize and the data finally become nationally representative data with good validity and reliability.

Thank you, both. let's say I have a variable that measures education from "before high school", "high school", "some college degree" ... "higher education".

Can I categorize the first two to one category and the remaining to another one? and then all the "missing data" and "not ascertained" to one variable representing missing data?

To understand your example, you have three variables originally. Without the “missing/not ascertained” in the initial categorization. You will want to recategorize to be able to accurately recode. Putting “before high school“ and “high school” in a box will depend on what analysis you are running in the recoded instance. If the aim is to concentrate on pre and post higher education that makes sense.

Yes it's possible. But pay attention to the characteristics (metric versus non-metric) of the various variables obtained and their adequacy to the requirements of the statistical methods to be used.

By recoding, you are lessening the variation in your independent variables. That may result in some loss of predictive power if that variation is important. On the other hand, if the two categories appear similar and you have only a small number of observations taking those values, then recoding may help to give you enough cases to reach statistical significance..

Dear friend

The following link can be useful.

https://libguides.library.kent.edu/spss/recodevariables

for descriptive narration recoding of categorical data not recommended. But for testing significance or association some times merging of categories is advised to make tests applicable and satify assumption of the tests.