I am about to validated a social cohesion scale, with a relatively large sample size (about 6000~). Aside from construct validity, I want to ask whether the following is acceptable as a measure of criterion validity. As social cohesion is a mulitlevel construct, I would like to see if social cohesion predicts national metrics like national crime rates or other social capital indicators from external data sets. I am uncertain because I would be using different sets of data as part of the validation process, and I am uncertain if this is accetpable in terms of validating scales.

Or in simple terms, must ALL the data used come from the SAME sample.