One more thing, the margin of error at the 95% confidence level is 7.5%. This means that if you asked a a random sample of 100 witnesses and ex-witnesses you would find that 56 to 70 of them had been born into the religion and then left as adults.
What does this mean? A larger sample size is needed.
While a larger sample size is always better, I stand by this as a decent survey. It's quite large at 35,000. Additional reasons why I like this is that it seems to correspond with approximately the percentage of missing or unaccounted baptized JW's which can be inferred from the data of baptisims vs publishers, previous statistics on disfellowshipping and disassociation, and population/death ratios. And, unfortunately, this is about as good as we are likely to get unless some researcher launches an independant statistical study of Jehovah's Witnesses with a budget big enough to match. One other interesting consideration is that the WTS clearly has the actual statistics (though maybe not any data on the success of 2nd generation vs 1st generation converts) and is not publishing them. You can construct a casual argument that because they would publish anything that looked good for them, the data actually looks bad.