The Mystery of RAND's odd numbers (Guest Post)

This is a guest post by Paul Mullen, Consultant Economist, Austin, TX; former Economic Advisor to the UK Minister for Local Government, Minister for Housing & Secretary for the Environment:

I would love to believe RAND's estimated 8.2 million increase in ESI between September and March but my experience in commissioning similar polls suggests to me that it is not real but due to an unanticipated side effect of the poll methodology.

We know that there is a pretty large "churn" in insurance status. Some people will gain jobs with insurance, while others lose them. Some people retire, get sick, or die. But one would expect the number gaining and losing employer coverage would be about equal. RAND's 8.2 million increase is the difference between 14.5 million estimated to have gained employer sponsored insurance and 6.3 million estimated to have lost it (with an estimated 102.4 million having employer coverage in both polls).

RAND's theory is that the threat of penalties has forced some to take up employer health care that they had chosen to go without before. In recent years the number of people covered by employer sponsored insurance has been declining, probably due to the ever increasing cost. So why would this trend suddenly be reversed so dramatically? If the individual mandate were the cause, surely there would have been all kinds of reports about such a dramatic increased enrollment? We have not seen any yet?

My belief is that the reported increase may be due to the design of the poll. This poll is unusual in that it interviewed the same people twice, first in September and then again in March. This allows it to see how many of the sample actually change from one type of cover to another.  But its weakness is that inevitably somewhere between 5 and 10% of the people you interviewed in the first round can't be contacted again. Some may refuse to answer more questions, but more likely the phone number has been disconnected. RAND reports that 2641 people were interviewed in September but only 2,425 of these (91.8%) were interviewed in March. So RAND went ahead with those who answered both surveys, assuming that were similar to the whole population.

Unfortunately this assumption is only justified if there is no correlation between life events that cause loss of employer insurance and RAND's failure to contact the sample again.  But why do phone numbers get disconnected?

• some people die or get sick and move into care;
• others move to a different part of the country (too far to keep their old number);
• some can't afford to pay their bill, or decide they can't afford to pay for both cell and home phone.

And why do people lose employer coverage? Basically they either lost their job or they chose to leave it.  If you lost a job you are more likely to be short of money and unable to pay the phone bill.You may also have to move to find work, or move in with relatives. You leave a job either because you found a better one (in which case you likely still get employer cover) or often because your family are moving to another City. New jobs may impose a waiting period for employer coverage. So it is quite possible that half the people who lost employer coverage between September 2013 and March 2014 were missed by this survey, as they were among the 8.2% of the sample who could not be contacted again.

Until we see other evidence to suggest that the mandate has caused a massive increase in take-up of employer offered cover, I just don't believe RAND's numbers.