Myth Busters #1: Roemer’s Law

If you ask anyone who has studied health economics or health policy in the last 40 years, “what is Roemer’s Law?” they will each be able to tell you in an instant — “that means a built bed is a filled bed.”

Milton Roemer, MD, was a researcher and professor, mostly at UCLA, who spent a lifetime (he died in 2001) advocating for national health systems around the world. He was involved in creating the World Health Organization in 1951 and Saskatchewan’s provincial single payer system in 1953. His “law” was based on a single study he did in 1959 that found a correlation between the number of hospital beds per person and the rate of hospital days used per person. That’s it.  That is the whole basis for “Roemer’s Law.”

“A built bed is a filled bed.” This little bumper sticker slogan has been the foundation of American health policy for 40 years. Hundreds of laws, massive programs, thousands of regulations at the federal, state and local levels of government, all have been based on this slogan. It is the source of such concepts as “provider induced demand,” and has resulted in centralized health planning, Certificate of Need regulations, managed care, and everything else currently on the table. Yet this “law” is both verifiably untrue and illogical.

There is a kernel of truth to it. When third-party payers pick up the tab, the usual tension between buyer and seller doesn’t exist. The buyer has no reason to resist excessive prices if someone else pays the bill.

But the believers in Roemer’s Law take that core idea to Alice-In-Wonderland proportions. They argue that, therefore, whenever a health care provider wants to make more money, it simply has to sell more services, and it will sell as many services as it has the capacity to provide — more capacity equals more sales without end. So, the only way to reduce this endless consumption is to limit the capacity — place strict controls on the availability of services.

But the notion fails for several reasons:

  1. People are not eager to enter the hospital, even when the cost is zero. Hospitals are miserable places to spend time. Folks are not lined up around the corner just waiting for an opportunity to be admitted to the hospital if only there were more beds available.
  2. If the “law” were true, hospital occupancy should approach 100 percent at all times. In fact, occupancy rates vary considerably over time and from place to place. Some years they are up, other years they are down.  For example from 1970 to 2000 national hospital occupancy rates dropped from 77 percent to 67 percent, according to the National Center for Health Statistics (page 368)  apparently one-third of “built beds” were not “filled beds” during this period. In 2005 occupancy rates varied from 92 percent in Delaware to 53 percent in Idaho.

So, Roemer’s Law is statistically untrue, it is behaviorally untrue, yet it has been the basis for virtually all of the health policy initiatives of the last 40 years, including Certificate of Need, national health planning, hospital rate-setting, Health Maintenance Organizations, and more recently, Accountable Care Organizations, pay-for-performance, and comparative effectiveness research.

No wonder all these efforts were doomed to fail — they were all based on a false premise, as we will look at in future installments.

Comments (11)

Trackback URL | Comments RSS Feed

  1. Devon Herrick says:

    Greg makes a good point. If supply induces its own demand, then hospital occupancy would approach 100% — which of course it doesn’t. I assume the relationship between supply and utilization varies by type of service. I can see how a physician-owned lab or MRI might boost spending since the equipment owner has the power to induce demand. But the solution is not Certificate of Need regulations.

  2. Linda Gorman says:

    What they’ve missed is that a built bed is not a staffed bed. Hospitals routinely adjust staffing to meet patient load but understand that adjusting capital is a lot harder. So hospital owners may build for projected peak patient load (depending on demographics in the area, projected growth, and likely payment capability) and staff for usual.

  3. Ralph Kristeller says:

    The more I read, the more I tend to believe that Economists know nothing about markets.

  4. Frank Timmins says:

    Geez, what foolishness. Do we even need statistics or behavioral studies to confirm that “Roemer’s Law” is wrong. As Greg suggested, who would…… (Let’s put it another way). What percentage of the population wants to spend any more time in a hospital room than they absolutely must?

    This is the most convoluted logic imaginable. Even though I have always heard of the health cost theories revolving around the notion that additional hospital beds invite over utilization, I did not know it evolved from the good doctor Roemer. Thanks Greg.

  5. Greg Scandlen says:

    Right, Linda. But that suggests that hospitals are responding to demand, not creating it.

    Devon, it has always perplexed me that, at the time the Stark laws were passed (restricting physicians from referring patients to labs they owned), one of the big scandals was the poor quality coming out of independent labs. What was driving physicians to own the labs was concern about quality control, not greed. Pete Stark was projecting his own dishonor onto doctors. And the idiot AMA went along with it.

  6. Linda Gorman says:

    Ralph. Please. As John Goodman has taken pains to point out, it is Health Economists, a subset of the entire profession, who seem to misunderstand markets. And that may be due to the subject, as he has also asserted that talking about health policy does bad things to people’s otherwise normal IQ.

  7. Eric says:

    How would you account for data such as this? (Figure 1).

    Basically, it shows that there is a strong linear correlation between the per-capita number of hospital beds and the number of hospitalizations for non-surgical conditions without clear guidelines (as compared to hip fractures, where there is no correlation). Obviously correlation is not causation, but I’d be curious to see what your take on these data would be.

  8. Virginia says:

    I had always suspected that there was excess demand for hospitalization prior to the hospital boom in the 1940’s and that growth in utilization was a function of this demand (coupled with an increase in technology the fueled more inpatient admissions).

  9. Frank Timmins says:

    @ Virginia

    When you say “excess demand” and “growth in utilization” does that mean “unnecessary” demand and “utilization” or simply that people were suffering medical conditions due to scarcity of hospital availability in the years after the depression?

    If hospitals were built to meet the demand of people who wanted to live longer, it would seem to have nothing to do with the notion that more hospital beds “create” the demand for hospital services.

  10. Greg Scandlen says:


    Thanks for the opportunity to comment. As you say — correlation is not causation — so let’s think about what the possible causes may be and certainly are not.

    1. The variation cannot possibly be caused by the payment system, since the payment system is identical in all the areas.

    2. The variation is unlikely to be caused by physician preferences, unless Dartmouth can demonstrate how physician training and/or culture varies by a similar factor from place to place. Curiously, Dartmouth has never even tried to do that. It has never even ventured a theory on it.

    3. What other factor might be involved? Dartmouth absolutely refuses to look at differences in patient demand and expectations. See my write up of its recent “study” on end-of-life care at —
    To the extent Dartmouth examines this at all, they confine themselves to variables like age, gender, and disease severity. They completely ignore factors such as living conditions, family support, religious orientation, education, or a host of other factors that go into making a complete human being.

    How well would I know you if all I knew was your age, gender, and disease severity?

    Now, let’s speculate about why Dartmouth has these blinders on. I suggest it is because Wennberg, Fisher, and the rest are so steeped in the Ivy League, elitist culture that they can’t imagine that an average working schlup would be able to influence the decisions of a well-trained clinician. I might be wrong, but it is the only explanation I can come up with.

    BTW, NCPA will shortly be posting an analysis I did of Wennberg’s original work comparing hysterectomies in Lewiston, ME and Wiscasset, ME. He did the exact same thing there — completely ignored that the women in Lewiston were almost entirely Roman Catholic and were demanding hysterectomies as an acceptable form of birth control in the 1970s.

  11. John R. Graham says:

    Eric & Greg: I ran a correlation between the number of jars of of pablum produced by Gerbers and the number of babies born. What do you know? A nearly perfect correlation!

    The government needs to limit the volume of baby food produced in order to cure the overpopulation problem.

    But it gets worse: I then ran a correlation between the number of diapers produced by Huggies and the number of babies born. Guess what? Yet another nearly perfect correlation!

    I then ran correlations with a whole bunch of baby products produced by very big companies like Procter & Gamble, Johnson & Johnson, et cetera. They all correlated very positively with the number of babies born.

    Clearly, we are facing a crisis of well-capitalized corporate interests that are conspiring to create “supplier-induced” demand for more babies!