If you can compute some numerical measure of an economy or society but it will clearly mislead people rather than inform them, don’t publish that indicator or cite it as evidence of anything. That’s not Ph.D.-level research methodology; it goes back to integrity and common sense. The corollary, of course, is that computing meaningful metrics often is far more complex than people realize.
Both those lessons are evident in the recent brouhaha over reported increases in health insurance rates offered through MNsure, our state’s exchange established under the federal Affordable Care Act.
MNsure administrators ignored that first rule and misled citizens by calculating and publicizing an unweighted average of premium increases for insurance companies that will participate in 2015. Their 4.5 percent “average” increase was a serious misrepresentation of what was really happening. Nearly all participants in the plans faced much larger increases, and I am not the only person who wondered, “Are these people crooked or just stupid?” on reading the details.
Jeff Johnson, the GOP candidate for governor, and others in his party are entirely correct in calling the 4.5 percent “bogus.” This is an egregious situation in which someone at MNsure or the Department of Commerce ought to be “seeking other opportunities in the private sector.”
Members of the public evidently see this as an arcane debate over technical details. They should be paying better attention.
The case also is a useful one for econ teachers like me because it embodies some of the same issues that apply to national economic indicators, especially price measures like the Consumer Price Index.
Start with a simple analogy to what MNsure did: Suppose 100 people in three groups start a diet. The two people in group A each lose four pounds. There are six people in group B and each gains three pounds. Group C has 92 members and each gains 10 pounds. Now figure the average weight gain or loss. You can do the math: -4+3+10=9. Divided the total by three groups and you find an average gain of three pounds.
But is that computed “average” gain of three pounds an accurate representation of what really happened, given that 92 out of 100 people gained more than three times this “average?”
According to MNsure’s logic, it is. But the aggregate change in weight for the 100 people was 930 pounds or an average of 9.3 pounds each. This is a far cry from three.
MNsure and the Commerce Commission are way off base and someone ought to be called to account. It would have been far more honest to report the weighted average premium inflation of 11.8 percent for the four plans that will remain on the exchange in 2015.
But these officials are correct in pointing out that it is impossible to calculate an exact measure of health premium changes. The primary reason is that there are announced new rates only for plans still in the exchange. But one plan, Preferred One, which had some 60 percent of participants in 2014, is dropping out because it misestimated costs badly. It cannot be clear how large the increases its marooned participants will experience as they move to other plans. These boosts will depend which plan they choose from which provider. That is unknowable right now. So any estimate of the cost increase computed over all 2015 participants will be imprecise. But clearly it will be much greater than 4.5 percent.
Note also that the percentage increases reported by the four remaining companies still participating are not across-the-board increases that are identical for every customer. Each insurer offers a tier of plans and different rates for different geographic areas. It is not even clear how carefully the overall percentages reported by each firm were weighted by sub-group. It is always dangerous to average averages, and here may they be doing it twice.
That isn’t the only caveat. In our example of weight gain, we assumed that the same 100 people were measured at beginning and end. But that clearly isn’t true for a health exchange involving more than 50,000 buyers.
Some who bought on the exchange in 2015 got jobs that offered coverage as a benefit, so they won’t be coming back. Some hit age 65 and went on Medicare. Some with employee benefits in 2014 lost jobs and are needing the exchange for the first time.
Young people who had been on their parents’ plans hit age 26 and, now on their own, are buying for the first time. Other healthy young people may decide it is cheaper to pay the penalty and forego coverage. Still others were burned by going naked and now seek coverage. Some people who had bare bones plans decided to get broader benefits and others are moving down. So there are myriad reasons why the premium paid by any individual goes up or down in the new year.
Moreover, the plans are making changes in the exact mix of covered and excluded services in each of their plans. New medical procedures and drugs get introduced all the time, and old ones are abandoned. So the products are not exactly the same.
If you see your monthly cost triple because you move from a bare-bones plan to a deluxe one, are you the victim of health insurance cost inflation? If you moved from a “gold” plan down to a “silver” one, are you the beneficiary of falling prices? If we tried to compute an equivalent to the 930-pound aggregate weight increase in our example, and tabulated the total increase in outlays for all participants, how much of this would be due to price inflation and how much to a different mix of participants and changes in products?
Answers to such questions are impossible to parse exactly. Yet these are precisely the challenges faced by the federal Departments of Labor and of Commerce as they tabulate inflation based on the CPI and other price indexes.
If consumers buy fewer cupcakes and more macaroons, how do you weigh things? If people respond to rising pork prices by eating more chicken, do you weigh the price changes by the relative 2013 or 2014 fractions of total household food spending that goes for each? And how much worse off is someone if they have chicken one more time each month and chops one fewer time? What about people who replace an old flip cell phone with a new 4G smartphone? Did their cost of living go up sharply or are they choosing a more luxurious pattern of consumption, no different from going to Cancun instead of Bemidji?
What is the real change in our “cost of living” if we spend $30 less per week on gasoline but $120 more per month on health insurance? And would it be accurate to combine a 5 percent increase in table salt with a 13 percent decrease in gas prices and say that inflation is 8 percent?
There is no right answer to such questions. Many people think the reported price indexes are rigged on direct orders from the Oval Office or that any proposed change, such as using a “chained” version of the deflator that does a better job of accommodating different consumption patterns, is a conspiracy to defraud the public. Economists know that, while not perfect, U.S. price statistics are about as accurate in measuring general price levels as one could get. Ineptitude in measuring prices by units of Minnesota government shows how bad the alternatives might be.