My co-authors, Coleman Drake and Sih-Ting Cai of the University of Pittsburgh and Dan Sacks of the University of Indiana, and I have a new working paper up at SSRN:
Zero-Price Effects in Health Insurance: Evidence from Colorado
This is work funded by the National Institute of Healthcare Management (NIHCM) which funded my and Coleman’s time. Connect for Health Colorado (C4HCO), the Colarado state based marketplace supplied us the data. All opinions and errors are ours alone and neither NIHCM’s nor C4HCO.
Our question was simple: Is Zero the hero when buying ACA health insurance?
Slightly more complexly, do people respond by buying zero premium plans way more than they respond if the cheapest plan that they see in their choice set is just a few cents or a few dollars more than zero?
We have good theoretical reasons to suspect that this could be the case. We have good theoretical reasons to suspect that this is not the case. Coleman and I had published a study in Health Affairs last January that used aggregate county level data to see if there was a zero premium effect. We found a 14% enrollment increase for the 151-200% FPL population in counties of states using on Healthcare.gov states where there was a zero premium plan available. The weakness of the Health Affairs study is that it used aggregated data and we could not see household responses. We had this result in mind when we wrote and received the NIHCM grant to look at specific household level response. My prior is that zero would have been special.
Zero premium plans occur on the ACA marketplaces when a chosen plan is composed only of EHB benefits and the premium spread between the chosen plan and the benchmark is bigger than the expected household contribution. Zero premium plans can only be plans priced below the ACA benchmark. Zero premium plans have always been fairly common for low income individuals who earn under 150% FPL as their expected contribution is small and the premium spread between the cheapest Bronze and the benchmark silver tends to be large. Very few people choose a zero premium plan in this income group; instead they go to high actuarial value CSR silver plans. Zero premium plans proliferated in 2018 to present due to the implementation of Silver Loading after the termination of direct federal reimbursement for silver plan CSR benefits. Bronze and Gold plans became relatively cheaper compared to silver plans in states that Silver Loaded. Colorado Broad Loaded in 2018 and Silver Loaded from 2019 to present. This gives us a ton of useful variation that enabled us to use regression discontinuity designs to poke at this problem.
So what did we find?
Did people buy Zero Premium Plans if they were available?
- Yes but this was part of the behavior of a significant chunk of people buying the cheapest plan available without regard to the price. Zero was not special compared to $1 or $5 or $10
Did people in Zero Premium Plans stay enrolled to December 31st of the year more frequently?
- Not really, compared to people who bought a cheap but not Zero plan, December 31st enrollment (and September enrollment as well) was about the same. We had hypothesized that duration would be different because it is hard to terminate someone for non-payment of premium when there is no premium to pay. Let’s put a pin into that
Did people in Zero Premium Plans have more duration?
- Why yes they did! ~50 days more enrollment/year.
How is that happening?
- Now this is the cool and interesting thing. People who had a zero premium plan exposure were far more likely to start coverage on January 1st than people who only had cheap but not zero premium exposure.
- There is a three step process to start a plan: Pick a plan, pay the initial binder payment, effectuate the plan. Plans that start on January 1st of the year have to be chosen by the end of OEP (usually December 15th or a couple of days after that if there is an extension). Now once a plan is picked, the potentially insured individual has to get a payment into the insurer before the plan year begins. Once the payment is received, the plan effectuates.
- People who have zero premium plans really only have a two step process: Pick a plan –> effectuate the plan
- People who have a non-zero premium plan can get dropped in the administrative/bureaucratic process of setting up payment.
How does this finding square with your Health Affairs paper?
- We think that the difference in Open Enrollment Period (OEP) structures matters a lot.
- OEP on Healthcare.gov for most of the relevant time period of increased zero premium plan availability ends on December 15th. Someone picking a non-zero premium plan on the 15th of December has to get the payment in on their first try or they never effectuate.
- OEP on C4HCO runs through January.
- An individual who picks a non-zero premium plan has multiple opportunities to “save” their enrollment. If they fail in December, they can try again in January to pick, pay and effectuate for either a February 1st or March 1st start date (depending on when in January they succeed).
- Our Health Affairs paper was looking at a marketplace that did not have these save opportunities which gave us the bump in enrollment from zero premium plans.
Why does this matter?
Zero premium plans likely don’t draw in new people from the margin compared to a $5 or $10 minimum premium plan. There is nothing special about zero in a price elasticity manner.
Zero premium plans increase enrollment duration because they meaningful reduce administrative burden and friction for low to moderate income buyers. From a policy point of view, this research has made me change my mind. I had thought that an OEP that ended in December versus one that ended in January was no big deal. I think I was wrong and I support either an extended OEP to at least January 15th OR a very permissive “save” period where people can go back and fix their errors and arrange payment for policies to start on February 1 if they had just messed up their binder payment.
Every step that is added to a process will increase the cost of completing the process. Administrative and process management costs are more significant than the premium costs in terms of enrollment and duration.
Finally, it was a blast to write and learn with and from Dan, Sih-ting and Coleman. This is the 8th piece I’ve written with Coleman (2 Health Affairs, 1 JAMA-IM, 1 JAMA Forums, JGIM, Psychiatric Services, The Conversation and now this working paper) and the 3rd wth Sih-Ting. Learning far more advanced methods from Dan Sacks was an eye-opener and a head twister. I am lucky that I get to learn and share new things with great co-authors.