By Milton Friedman
The Wall Street Journal
Some years ago, I came across a study by Max Gammon, a
British physician who also researches medical care, comparing
input and output in the British socialized hospital system.
He took the number of employees as his measure of input and
the number of hospital beds as his measure of output. He
found that input had increased sharply, while output had
He was led to enunciate what he called "the theory of bureaucratic displacement." In his words, in "a bureaucratic system . . . increase in expenditure will be matched by fall in production. . . . Such systems will act rather like `black holes,' in the economic universe, simultaneously sucking in resources, and shrinking in terms of `emitted production.'"
I have long been impressed by the operation of Gammon's law in the U.S. schooling system: Input, however measured, has been going up for decades, and output, whether measured by number of students, number of schools, or even more clearly, quality, has been going down.
The recent surge of concern about the rising cost of medical care, and of proposals to do something about it -- most involving a further move toward the complete socialization of medicine -- reminded me of the Gammon study and led me to investigate whether his law applied to U.S. health care. There clearly have been major advances in medical care in the past half century. Indeed, I would not myself be alive today if it were not for some of them. Yet the question remains whether these gains were promoted or retarded by the extraordinary rise in the fraction of national income spent on medical care. How does output compare with input?
Even a casual glance at figures on input and output in U.S. hospitals indicates that Gammon's law has been in full operation for U.S. hospitals since the end of World War II, and especially since the enactment of Medicare and Medicaid in 1965.
Before 1940, input and output both rose, input somewhat more than output, presumably because of the introduction of more sophisticated and expensive treatment. The cost of hospital care per resident of the U.S., adjusted for inflation, rose from 1929 to 1940 at the rate of 5% per year; the number of occupied beds, at 2.4% a year. Cost per patient day, adjusted for inflation, rose only modestly.
The situation was very different after the war. From 1946 to 1989, the number of beds per 1,000 population fell by more than one-half; the occupancy rate, by one-eighth. In sharp contrast, input skyrocketed. Hospital personnel per occupied bed multiplied nearly seven-fold and cost per patient day, adjusted for inflation, an astounding 26-fold. One major engine of these changes was the enactment of Medicare and Medicaid in 1965. A mild rise in input was turned into a meteoric rise; a mild fall in output, into a rapid decline.
Taken by itself, the decline in the number of occupied beds could be interpreted as evidence of the progress of medical science: The population is healthier, needing hospitalization less, and advances in science and medical technology have reduced the length of hospital stays and enabled more procedures to be performed outside the hospital.
But that does not explain much, if any, of the rise in input. True, care has become more sophisticated and expensive, and medical machines more complex. Yet improvements in health and hospital care quality do not appear to have proceeded more rapidly after 1965 than before -- there is some evidence that the reverse was true. Reported expenditures on research (per capita and in constant dollars) rose at the rate of 15% a year from 1948 to 1964, at less than 2% a year from 1965 to 1989. Yet the number of occupied beds per 1,000 population fell 1% a year from 1946 to 1964, and 2.5% a year from 1965 to 1989. Cost per patient day rose by 6% a year in the first period, 9% in the second.
Gammon's law, not medical miracles, was clearly at work. The federal government's assumption of responsibility for hospital and medical care for the elderly and the poor provided a fresh pool of money, and there was no shortage of takers. Personnel per occupied bed, which had already doubled from 1946 to 1965, more than tripled from that level after 1965. Cost per patient day, which had already more than tripled from 1946 to 1965, multiplied a further eight-fold after 1965.
Growing costs, in turn, led to more regulation of hospitals, further increasing administrative expense. Unfortunately, I have been unable to uncover comprehensive and readily available data for a sufficiently long period to judge how large a role was played by increasing administrative costs. Anecdotal evidence suggests that increased administrative complexity played a major role in the explosion of total cost per patient day, and led to a shift from hospital to outpatient care, accelerating the decline in occupied beds.
Experts in medical care and in hospital administration can doubtless put flesh on the stark evidence from the limited statistical data. But a fuller description is hardly likely to alter the bottom line: In Gammon's words, "a bureaucratic system . . . will act rather like `black holes,' in the economic universe."
Though hospital cost has risen as a percentage of total medical cost from 24% in 1946 to 36% in 1989, it is still a minor part of total medical cost. It is tempting to apply Gammon's analysis to total medical cost rather than simply to hospital care.
There is no problem about input. Estimates of expenditures on medical care are readily available for the postwar period, can be estimated back to 1919, and can be corrected readily for the rise in population and in the price level.
Except for the Great Depression, when the collapse of incomes raised the percentage sharply, health spending from 1919 on rose gradually but stayed between about 3% and 4% of total national income. Government spending was only a modest part of that total and was primarily state and local. After the war, total spending on health care tripled as a fraction of national income; government spending, particularly federal, became an increasing fraction of the total.
Private spending rose at a steady arithmetic rate up to the end of World War II, increasing by $3.30 per capita a year in 1982 dollars, with only minor deviations as a result of cyclical forces. As a percentage of national income, private spending stayed between 3.5% and 5% from 1922 to 1958 except for some of the Depression years. From 1958 on, private spending began to rise as a percentage of national income -- at first slowly, then more rapidly -- reaching more than 8% by 1989.
Government spending behaved somewhat differently. It rose at a rather constant percentage rate from 1919 all the way to 1965, except for a short postwar bulge. The enactment of Medicare and Medicaid produced an explosion in government spending. In the process, government's share of total spending went from 15% during the 1920s to 25% in 1965, before surging to 42% in the next two decades, or from less than 1% of national income to nearly 6%. If the prior trends had continued, total spending in 1989 would have been about half as much as it actually was.
One major physical input is the number of physicians. Physicians numbered 157 per 100,000 population at the turn of the century, gradually declined in number to 125 by 1929, and then rose slowly to 133 by 1940 before beginning an exponential climb to 252 by 1987, the latest year for which I have data. The rapid increase in the number of physicians was preceded by a sharp rise in their median income from a level less than seven times per-capita income to a peak of 11.6 to 1 in 1962. As cost containment became more and more pressing, the rise in the number of physicians was accompanied by a decline in their relative income, even though their income continued to rise in absolute terms. By 1987, the ratio had declined from 11.6 to 9.1, and no doubt the decline is continuing.
Despite the sharp rise in the number and income of physicians, it is worth noting, first, that the cost of physicians' services accounts for only about one-fifth of total health-care costs, and, second, that the share is less than it has historically been. In 1929, the cost of physicians' services was about 27% of total health costs; after World War II, about 25%. The explanation is presumably a combination of more expensive equipment and greater administrative expenses.
So much for input. What about output? That is the real problem. The output of the medical-care industry that we are interested in is its contribution to better health. How can we measure "better health" in a reasonably objective way not greatly influenced by other factors?
The least bad measure that I have been able to come up with is length of life. That too is seriously contaminated by other factors. Improvements in diet, housing, clothing and so on made possible by increasing affluence as well as such government measures as the provision of purer water and better garbage collection and disposal have doubtless contributed to lengthening the average life span. Wars, epidemics, and natural and man-made disasters have played a part. Even more important, the quality of life is as important as its length. Perhaps someone more knowledgeable in this field can find a better measure of the relevant output of the medical-care industry. I have not been able to.
Data on life expectancy at birth are readily available by sex and race, and I have concentrated on the length of life of females, and of whites and blacks separately, in order to keep the populations as homogeneous as possible over a long period. Life expectancy at birth of white females rose steadily from 48.7 years in 1900 to 74.2 years in 1959, and of black females from 33.5 years to 65.2 years. The rise then slowed drastically. Life expectancy went from 74.2 in 1959 to 79.0 in 1989 for white females and from 65.2 years to 75.6 years for black females. The rate of rise was more than halved for whites, cut by more than a third for blacks.
As the length of life increases, further increases are presumably more and more difficult to achieve -- early gains would seem to be the easiest. Yet there was no sign of any slowdown for the first 59 years of the 20th century. The shift to a lower rate of improvement comes suddenly, not long before the rapid expansion in the federal government's role in medical care and the sharp slowdown in the rate of increase in the funds going to research.
Life expectancy at age 65, in sharp contrast, shows very slow though steady progress to about 1939, and then decidedly more rapid progress, especially for females. Does the speeding up around 1939 reflect the discovery about that time and subsequent wider use of a widespread range of antibiotics? I leave that as a puzzle for others.
In terms of my own concern, the effect of greater government involvement, the data are of little help. For females, Medicare is followed by an initial speeding up, then tapering off; for males, the pattern is almost the reverse: little or no change from 1950 to 1970, then a speeding up. In short, it will take a far more detailed and informed analysis to reach any clear conclusions about what has been happening to the output of the medical-care industry, whether in terms of the length of life, and even more, the quality of life.
Nonetheless, for total medical care, as for hospitals, it is hard to avoid the conclusion that Gammon's law is at work. No question that medicine in all its aspects has become subject to an ever more complex bureaucratic structure. No question that input has exploded. No evidence that output has come anywhere close to keeping pace, though we lack a firm basis for going beyond this very general statement. "Black holes" indeed.
Why should we be surprised? Evidence covering a much broader range of activities documents that bureaucratic structures produce high cost, low quality, and inequitable distribution of output. The U.S. medical system has become in large part a socialist enterprise. Why should we be any better at socialism than the Soviets? Medicine is not unique. Our socialized postal system, our socialized schooling system, our socialized system of trying to control drugs, and indeed our socialized defense industry provide clear evidence that we are no better at socialism than countries that have gone all the way.
Yet not only do we keep on being surprised, but we continue in each of these areas to increase the scope of socialism. Nearly all the suggestions for improving our medical system involve expanding the role of government, at the extreme moving from a partly socialist system to a completely socialist system!
The inefficiency, high cost and inequitable character of our medical system can be fundamentally remedied in only one way: by moving in the other direction, toward re-privatizing medical care. I conjecture that almost all consumers of medical services, and many producers, would favor a simple reform that would privatize most medical care. Yet that reform is politically infeasible because it would be violently opposed by the bureaucracy that plans, controls and administers today's structure of medical care.
The reform has two major steps: (1) End both Medicare and Medicaid and replace them with a requirement that every U.S. family unit have a major medical insurance policy with a high deductible, say $20,000 a year or 30% of the unit's income during the prior two years, whichever is lower. (2) End the tax exemption of employer-provided medical care; it should be regarded as a fully taxable fringe benefit to the employee-deductible for the employer but taxable to the employee. Each of these reforms needs further discussion.
Preferably, the major medical insurance policy should be paid for by the individual family unit, which should receive a reduction in taxes reflecting the reduction in cost to the government. There would be an exception for lower-income families and for families who were unable to qualify for coverage at an affordable fee. The government would help them finance the policy though not administer it. That would be done by private competitive insurers, chosen by each individual or family separately. Each individual or family would, of course, be free to buy supplementary insurance if it so desired.
However, even if the government were to pay directly for major medical insurance for everyone -- rather than by reducing taxes -- there is little doubt that both the government's and the total health cost would decline drastically because of the elimination of the tremendous governmental bureaucratic structure that has been built up to supervise a large fraction of all health activities.
The tax exemption of employer-provided medical care has two different effects, both of which contribute to raising health costs. First, it leads the employee to rely on the employer rather than himself to finance and provide medical care. Yet the employee is likely to do a far better job of monitoring health-care providers in his own interest than is the employer. Second, it leads him to take a larger fraction of his total remuneration in the form of health care than he would if it had the same tax status as other expenditures. If the tax exemption were removed, employees could bargain with their employers for higher take-home pay in lieu of health care, and provide for their own health care, either by dealing directly with health-care providers or through purchasing health insurance.
These two reforms would completely solve the problem of the currently medically uninsured, eliminate most of the bureaucratic structure, free medical practitioners from an incredible burden of paper work and regulation, and lead many employers and employees to convert employer-provided medical care into a higher cash wage. The taxpayer would save money, since total governmental costs would plummet.
The family unit would be relieved of one of its major concerns -- the possibility of being impoverished by a major medical catastrophe -- and most could readily finance the remaining medical costs, which I conjecture would return to something like the 5% of total consumer spending that it was before the federal government got into the act. Families would once again have an incentive to monitor the providers of medical care and to establish the kind of personal relations with them that once were customary. The demonstrated efficiency of private enterprise would have a chance to operate to improve the quality and lower the cost of medical care.
There is only one thing wrong with this dream. It would displace and displease the large number of people who are now engaged in administering, studying and daily revising the present socialized system, including a large private-sector component that has adjusted to the system. They are sufficiently potent politically to kill any such reform before it could ever get a real following, just as the educational bureaucracy has repeatedly killed even modest programs for privatizing the educational system, even though poll after poll shows that the public supports privatization through parental choice.
Medical care provides a clear example of the basic difference between private and governmental enterprise. The difference is not in the quality of people who initiate or operate new ventures, or in the promise of the ventures. The difference is in the bottom line. If a private venture is unsuccessful, its backers must either shut it down or finance its losses out of their own pockets, so it will generally be terminated promptly. If a governmental venture is unsuccessful, its backers have a very different bottom line.
Shutting it down is an admission of failure, something none of us is prepared to face if we can help it. And they need not shut it down. Instead, in entire good faith, the backers can contend that the apparent lack of success is simply a result of not carrying the venture far enough. If they are persuasive enough, they can draw on the deep pockets of the tax-paying public, while replenishing their own, to finance a continuation and expansion of the venture. Little wonder that unsuccessful government ventures are generally expanded rather than terminated.
In my opinion, that is what is responsible for Gammon's "black holes," whether in medicine, schooling, the "war on drugs," agricultural subsidies, protectionism, and so on and on. That is the way high-minded motives and self-interest combine to produce what Rep. Dick Armey once labeled "the invisible foot of government." I challenge you to find more than a very exceptional counter-example.
Mr. Friedman, a Nobel laureate in economics the year this was published, is a fellow at the Hoover Institution.
Gary S. Becker, Robert J. Cihak, James F. Fries, Thomas Moore and Gloria Valentine helped this article through their work and comments.