Very practical but tough question to government data producer organizations . I think there should be no objection in using published, publically available digital databases as a source of data to answer questions surrounding policy decisions. However, "Are we able to calculate error with any confidence"? is a big question mark. In my opinion the answer to this part is NO.

I agree with Asmat.

On the other hand, you can say that every dataset is not confident in 100%.

So, there is a common practice in research filed, that "something" was estimated rather than "something" is equal.

Any available statistics are just estimation of real level of "something" (e.g. consumption, foreign investments). Nevertheless there are used by government agencies, central banks and scholars.

@Paweł Weichbroth, I agree to your argument and would like to add:

True value of "something" can never be known.

Consider that important policy decisions are made based on findings of research using databases that we have no idea of what the error is. Is the more pertinent question "should we use these databases?". The layman and policy makers are depending on the scientists to deliver reliable information on which to make policy decisions. If we are no able to define or estimate reliability, are we being responsible scientists?

Researchers having an academic affiliation and access to the primary data may serve as analyst for some investigations. However provided, such a data registry must be accessible to the public at no charge, searchable, open to all prospective registrants, and managed by a not-for-profit organization. The registry must include the following information: a unique identifying number, a statement of the intervention(s), study hypothesis, definition of primary and secondary outcome measurements, eligibility criteria, target number of subjects, funding source, contact information for the principal investigator, and key dates (registration date, start date, and completion date).