This is the interesting conclusion of a new paper, "Take-Up, Drop-Out, and Spending in ACA Marketplaces" by Rebecca Diamond, Michael Dickstein, Tim McQuade, and Petra Persson. One good summary graph:
The authors measure health spending such as copayments and premiums out of bank data. A good table:
The important numbers are in panel B, among the "middle class."
"Panel B shows that enrollees under the ACA exhibit large increases in health spending when covered by insurance. Consumers with annual income less than 20K, for example, increased their health spending by 28% when covered and then cut their spending by 40% after dropping out. (My emphasis) Those with annual income in the 20K-40K range increased spending by 54% when covered, and cut spending by 51% when dropping out.(Yes, the huge numbers <20 pre ACA are interesting. It's a much smaller sample though -- people with < 20k of annual income who bought private health insurance.)
So, how do insurers respond when relatively healthy people are dropping out, and only chronically sick people keep insurance? Don't jump too quickly to conclusions.
The authors compare counties with high dropout rates to counties with low drop out rates. If there is a lot of competition in health insurance, you'd expect high dropout rates to lead to bigger price increases, to cover costs. If not, however, low dropout rates are a sign of a lot of desperate people with chronic conditions who will pay no matter what the premium
Here is a regression across counties. Low dropout rate leads to higher prices.
The conclusion is pretty clear. If we're going to have health insurance with guaranteed issue -- same price no matter your health status -- dropping out must have some consequences. The paper computes the cost of the current penalty (not well enforced, I gather) and given the current ease of signing up again if you get sick, finds the penalty not effective.
Of course, we could transition to personal, portable, guaranteed-renewable, time-consistent insurance instead...Oh, stop dreaming.
A minor interesting tidbit. Their data consists of all transactions by customers of a major bank. Like this:
Et tu, banks? I didn't realize banks sold this kind of data. Though it does not have person identifiers, it would be easy enough to figure out my record. Or President Trump's.