My thoughts exactly, Leisuredude. Quick math for average family with $75K income would save us over $10K EVERY YEAR! What a huge stimulus to the overall economy THAT would be, as most folks like us would spend 90%+ of that extra spending money! That’s the point that Bernie hasn’t made explicit when he says you’ll save more in premiums, deductibles, and copays than your increased taxes, which scares some people. Just tell them everyone but the super rich gets a 14% permanent RAISE and MFA sells itself.

# Healthcare for Everyone and It Will Cost Less: The Economic Case for Medicare for All

Exactly. And as for who assumes the risk, that is ONE of the many reasons the call–if not from the beginning certainly as the subject has been more widely discussed in recent years–has been for EXPANDED and IMPROVED Medicare for all: For the system to work at all the risk must fall on the insurer, who in turn must have some transparent leverage over the actual providers. I have not yet read the bill (has it even been posted?), but if it merely transfers the risk to the providers that must be fixed.

Why focus on some modest savings instead of the real opportunity for savings, more efficient risk management?

A single payer, not for profit, health insurer is the most mathematically efficient risk manager available for insuring the entire population of the USA.

The costs of risk management in smaller insurers far exceed the costs of risk management in larger insurers, leading to failed insurers, decreased benefits, and excessive profits compared to the costs of risk management for a largest possible health insurer.

So let’s consider a simple case of two insurers, both selecting policyholders, at random, from the population of the USA and offering exactly the same benefits.

To get started, let’s assume there is a “big enough” health insurer, I will call the Paradigm Insurer. Our analyses will be derivative from the characteristics of PI and as a society we think PI is efficient enough to think that 327 health insurers will offer more than enough consumer choice and competition, though competition in insurance is inefficient as I will show.

PI insures 1,000,000 policyholders. PI’s loss ratio is a random variable, LR, which is an estimate of the population loss ratio (PLR = 0.7500) for the USA, Over many years of study we know that PI’s loss ratio, LR, varies around the population loss ratio, with a standard error = 0.05. In short, PI’s loss ratio lies below 0.7500 in about half of all years.

PI’s loss ratio lies belwo 0.8000 in about 84 out of 100 years, and PI’s loss ratio lies below 0.8500 in about 98 years in 100.

PI’s expense ratio is about 15% of premium revenue, and PI has a target profit ratio of 5% of premium revenues. Because PI’s loss ratio is a randome variable, it won’t alway be high enough to generate profits of 5%. PI charges a “Risk Premium” of 5% of revenues to improve its probability of earning profits and avoiding losses. The risk premium is a measure of insurer efficiency. The smaller the risk premium an insurer needs to meet its profit goals, the more efficient it is.

Based on these assumptions, PI can expect, at worst, to break even, at loss ratios less than 0.8500, with probability 0.9772. PI can expect, at worst, to earn profits of at least 5% of revenues, at loss ratios less than 0.8000, with probability 0.8413. PI can expect, at worst, to earn profits greater than 10%, at loss ratios less than 0.7500, with probability 0.5000.

On the downside, PI can expect, at best, to incur losses greater than 0%, at loss ratios above 0.8500, with probability 0.0228 (1.000 - 0.9772. PI’s probability of losses greater than 10% of revenues is negligible, at loss ratios in excess of 0.9000, which will occur with probability 0.0000.

I could detail a lot more characteristics, but this will suffice. These analyses are, of course, based on the normal distribution, which is acceptable because insurer sample sizes are huge. Some psychologists invoke normality with sample sizes as small as 8-10 subjects.

How about an insurer with only 100,000 policyholders, also selecting policyholders at random, and offering identical benefits? I’ll call this health insurer TS, for “Too Small”.

PI’s characteristics are all based on PI’s standard deviation, of 0.0500. There is actually a very long name for PI’s standard deviation, it is also called standard deviation of the sampling distribution of the sample statistic for a sample of size 1,000,000 when drawing policyholders at random from the population of the USA. You know this better as the “Standard Error”.

TS has the same probability of earning profits greater than 10% of revenues. This is why a lot of researchers go no further. They think all insurers have the same operating characteristics. But TS’s standard error is more than 3 times larger than PI’s standard error, or 0.1581 (sqrt(10) * 0.0500).

With such a large standard error, TS’s fortunes are very different than PI’s fortunes.

TS can expect, at worst, to break even, at loss ratios less than 0.8500, with probability 0.7365. TS can expect, at worst, to earn profits of at least 5% or revenues, at loss ratios less than 0.8000, with probability 0.6241. TS, like PI, can expect, at worst, to earn profits greater than 10%, at loss ratios less than 0.7500, with probability 0.5000.

Worse still, on the downside, TS can expect, at best, to incur losses greater than 0%, at loss ratios above 0.8500, with probability 0.2635 (1.0000 - 0.7365). PI’s probability of losses greater than 10$ of revenues is negligible, at loss ratios in excess of 0.9000, with probability 0.0000, but TS’s probability of losses greater than 10% of premium revenues is far higher, at 0.1030. TS will incur potentially catastrophic losses more than 10 years in 100.

I’ll leave you to do some more computing, or you can just read my book: “Standard Errors: Our Failing Health Care (Finance) Systems and How to Fix” https://www.standarderrors.orgs for all the answers and to see some more problems TS faces.

So now, let’s consider a single payer, national health insurer (NHI), with 323,000,000 policyholders. NHI has a really, really small standard error, just 0.00278, compared to PI’s 0.0500, and TS’s 0.1581.

NHI can expect, at worst, to break even, at loss ratios less than 0.8500, with probability 1.0000. With the same premiums per policyholder, and providing identical benefits, NHI always earns some profits. NHI can expect, at worst, to earn profits of at least 5% of revenues, at loss ratios below 0.8000, with probability… 1.0000. PI, because it is such an efficient risk manager, will always earn profits of at least 5% of premium revenues. Because NHI is so large, it has eliminated all risk of adverse operating results.

We saw earlier that PI’s probability of incurring losses greater than 10% of revenues at a loss ratio four standard errors higher than the population loss ratio (0.7500), was 0.000 because there is so little probability four standard errors above the PLR. But NHI’s loss ratio would be four of its standard errors higher than the PLR at a loss ratio of 0.76112 (0.7500 + 4 * 0.00278). NHI will earn profits greater than 8.89% of premium revenues every year because NHI is an incredibly efficient risk manager.

What can NHI do that PI and TS can never do? Offer higher benefits or charge lower premiums for the same benefits.

My book works out all the other details as well, including the fact that to match PI’s probability of earning modest profits of just 5% of premium revenues, TS would have to cut benefits drastically - which is exactly what health insurers, and insurance risk assuming health care providers do, in practice to avoid going incurring losses and potentially going bankrupt.