Sometimes it's better to split the baby

This story about the Census Bureau is amazing to me: The Census is changing its annual survey about health insurance.  As a result, the new data will not be comparable to the old, making it much harder to gauge the effects of the Affordable Care Act.

Is this a White House conspiracy to hide the effects of the law, as some have suggested?  Maybe, but probably not. I have a lot of respect for the government data producers, so I am giving them the benefit of the doubt.

Yet I don't see why the Bureau needs to make such a sudden change.  Why not, for a few years, give half the sample the old questionnaire and half the new one?  This procedure would provide a basis for eventually splicing together the old and new time series.