How accurate is that 2-3% attrition rate estimate, anyway??
I just posted an entry noting that Washington State was reporting having exactly 156,155 paid exchange-based QHP enrollments as of June 1st.
Now, this is noteworthy because according to an earlier update, as of around a week earlier, the state had reported adding "between 2,000 - 3,000" exchange QHP enrollees between the end of open enrollment on 3/31 (WA did not offer an extension period) and May 27th; let's split the difference and call it 2,500. Add that to the official 3/31 total and you get around 166,500.
What accounts for the roughly 10,000 person difference? Well, the first number includes both enrollments and cancellations after the first month. Remember, Washington State only reports enrollments once the first month's premium has actually been paid, so these should be "clean" numbers.
In other words, between 3/31 and 6/1, Washington added about 2,500 people. Meanwhile, throughout the entire open enrollment period, the WA exchange lost around 10,300 enrollments.
Assuming that the vast majority of these happened after 3/31 (which is likely, since 40% of total enrollment happened after 3/15), 10,300 out of 166,500 = an attrition rate of around 6.2% over 2 months, or around 3% per month.
As it happens, that's exactly within the 2-3% attrition rate (if at the high end) that anti-ACA analyst Robert Laszewski thinks sounds about right, and which I've accepted as seeming pretty reasonable. In fact, it's actually better than that, since some of that 10,300 drop includes people who dropped out earlier in the year (between February and March) as their circumstances changed, meaning that 2-3% does indeed sound like a pretty solid range to me.
To my knowledge, this is the first hard number we have which (kind of) directly compares additions vs. cancellations after the first month. While it's only one state and only runs through June 1st, it's still better than nothing, and it makes me feel better about using the 2-3% estimate going forward (at least until more data is provided, of course).