Not long ago TAI published (online and in the print magazine) an essay by Scott Alexander called “Notes on Cost Disease.” This essay originated as an extended blog post from the author and, like a lot of blog posts, it was written in a personal, loosely structured way. The author, whose real name I came to learn is not (just) Scott Alexander buried his lede about two-thirds of the way into the piece.
Moreover, his palate of explanations for the phenomenon he identified struck me, as an omnivorously curious but “non-practicing” (that is, not an academic or think-tanker doing original research) social science type, as probably incomplete. It turned out that the author is not, as many of his blogosphere followers assumed, an economist or a social scientist of another flavor, but a psychiatrist practicing “somewhere in the Midwest.” Maybe Detroit. So the paucity of explanation in the blog post ends up being not so surprising; one would want a medical doctor to be better and more deeply read in the medical literature than in decades’ worth of social science esoterica. Still, he had the very big right question, impressive data sets at least related to an answer, and an obviously formidable capacity for creative analysis, so I took the plunge: As I sometimes do, when I find an idea powerful and interesting enough, I will do whatever is necessary to repackage it properly and present it to readers.
The outcome of the effort has been well worth it: The essay is dynamite. I wish I could force it down the throat or inject into the bloodstream of every member of the American political class (and that’s what it would take to get the idea across since so few of them deep read anymore).1 Everyone who cares about the underlying condition of the American economy, and its social and political implications (which are vast…just look at who occupies the Oval Office) should read the Alexander essay, and I would wager that a substantial number of those who do will experience that frisson of excitement trilling the hairs up and down one’s neck. This essay is a Necker Cube flipper, a paradigm somersault maker, a brain lift of the first order.
Maybe the piece could have been still better. I asked the author to allow me to suggest via the edit a range of other explanatory possibilities that might give the essay greater heuristic power. He agreed to let me try but, as an honest man, he declined to accept some of these suggestions because he could not personally vouch for them. (He did take on others.) He is not, to repeat, a social science guy, anymore than I am a psychiatrist guy. So fine; I understood and let it pass.
Except I can’t let it pass completely. So let me now show you the omitted suggestions I made, and let me note as I do that none of these suggestions are assertions, merely research avenues that might bear a fruitful trip or two. I do this in a spirit of collegial aspiration: I would love nothing more than to stimulate some smart people, economists and others, to apply themselves to the mystery of cost disease to see if we, together, can come up with a better account of what is happening. If we can, we will have taken a first step, perhaps, to knowing what to do about it.
Before discussing explanatory possibilities not in the Alexander essay, let’s first review briefly those that are there—which the author freely admits do not sum to a persuasive answer.
First is the famous 1960s-era Baumol Effect, which Alexander raises, explains, and pretty much dismisses. The Baumol Effect is based on the simple idea that a significant productivity increase in one sector will raise many if not all boats in other sectors because labor pools are malleable over time and will follow high-wage work unless incentives are proffered to keep them busy elsewhere. Incentives are offered, says the Baumol Effect, and so wages (and costs passed along) rise even when no productivity increases justify them.
Alexander’s description of the Baumol Effrct is fine enough, so if you don’t already know what it is, just go back and read his summary. I, for one, have never been entirely persuaded by this argument because the fungibility of the labor pool, even over time, seems to me to be more constrained than the theory suggests. The original formulation contrasts high-paying manufacturing labor aided by technological innovation with work in a symphony orchestra, and Alexander repeats the pairing. I doubt, however, that a young music student—especially perhaps a female one—is likely to switch aspirations and go into a manufacturing career just because the pay figures to be higher. The old model of the individual, functionally interchangeable value-maximizer as the ideal actor model in economic theory never made much sense. The storied originators of this theory, mostly at the University of Chicago, knew its limits and acknowledged that it was just a model useful for certain purposes but not others. (Those who came after sometimes ignored the caveat about limits.) So if the Baumol Effect is real, the means by which it works are probably subtler than originally imagined and the maximal effects probably more limited. I wonder, too, how much empirical research exists on the Baumol Effect; I suspect not very much.
But, in any event, Alexander dismisses the Baumol Effect as a major source of cost disease because labor niches where major productivity increases have not occurred for one reason or another—health care, education, housing, and others—do not show sustained relative wage growth over time, in this case about fifty years. It may not be such a good idea to dismiss the Baumol Effect so quickly, however, of which more below.
Other explanations offered in the essay—some of them in the original blog and some of them my interpolations as editor—include the possibility of massive market failures in certain areas because of institutional rentier behavior, some of which may be further shaped by socially mediated factors. (Again, read the original, but note here that the idea of institutional rentier behavior differs from the classical definition, which focuses on oligopolistic behavior—so, for example, Carefirst is an institution heavily involved in health care, but as a service provider whose guard rails for profit are defined largely by government it does not fit the definition of a participant in a private industrial/manufacturing oligopoly. This is a distinction about which, I suspect, little analysis has been done thus far.)
They also include the cost of government ineptitude, but Alexander dismisses the standard libertarian argument here by noting that private hospitals and schools are not on the whole that much less vulnerable to cost disease than public ones. He focuses instead on regulation as a cost of government, and in particular on the real and anticipated costs of litigation. Housing is a particularly vivid example, but Alexander does not go into this in depth, possibly because it isn’t easy to separate out the cost disease effects of safely regulations as against the effects caused by land-use policy regulations. HIPAA is another example Alexander might have mentioned but did not: A whole private service industry has arisen in recent years—and not a small one—whose purpose is to guide medical practices through the maze of regulations pertaining to electronic medical records. As a concept, HIPAA is necessary in an age of digitized records, but no doctor I have ever met thinks the system the reg-writers created is anything other an unwieldy, counterproductive, and massively expensive albatross.
Yet Alexander also notes, quite shrewdly, that some of this regulation burden may be because social norms have changed, people being less willing to run risks of any kind but not able to reckon the wider social costs of indulging that new norm. So government is not entirely to blame in a democracy if more regulation is what people actually want to shield them from risk.
And lastly, Alexander points in the essay to the role of delinquency in cost disease—some people defaulting on payment, leaving others in the relevant pool liable collectively for their missing share. He mentions examples of delinquency from the worlds of student loans and health care, but other examples may be cited, too. It also applies to car insurance, for example.
In some auto insurance pools—like large cities with a lot of poor people—costs are much higher because a fairly large number of people go driving around without insurance (and often without even a valid license), even though doing so is against the law. But the city—Philadelphia is a stark case in point—for whatever reason, doesn’t enforce the law, so responsible consumers in the insurance pool pay the consequences when uninsured motorists cause harm to others. (One would think there would be more class-action lawsuits against city governments that act in such a way. There ought to be.) Just how much does delinquency in this case contribute to cost disease in auto insurance rates, or for that matter in student loan interest rates and health care costs? That would be a fascinating area of comparative research. I, for one, would like to see the results.
But even by Alexander’s own calculus, all these putative sources of cost disease do not add up to a satisfying explanation of his own data sets concerning health care, housing, infrastructure, education, and other economic domains. So what did I add that the author declined to take on?
I added one small idea about an economic oddity that is in some ways a variation of the Baumol Effect in the sense that it shows how market dynamics can have unexpected outcomes related to cost disease. And I added two larger suggestions, one that asks us to look again at the data concerning the Baumol Effect, and another that concerns the generic issue of transactional costs, which Alexander mentioned in passing in his blog but did not pursue. Let’s look at these in turn.
Cost-Pull Progress. Say there is a medical breakthrough of a sort that combines some new method of scientific measurement with a procedure (or a pharmacological application) that can remedy or sharply ameliorate some costly chronic ailment (an array of diabetes symptoms, say). And suppose that this new ensemble of capabilities, though not cheap, is nevertheless vastly cheaper than the rippling costs of dealing with a chronic ailment over an extended period. So we would have what amounts to a productivity advance for treating a disease, and for any given patient thus afflicted costs would go way down as reckoned over time as a result.
Ah, but the desirability of the new procedure brings a great many people to seek it, thus sharply increasing demand and pushing up the cost of applying the breakthrough. So, as in the Baumol Effect, we have a productivity gain that can nevertheless increase costs per unit of application of an innovation as well as aggregate spending (two different things2), at least for a time, instead of lessening the cost as normal economic logic would expect. (It could be hard to measure this because, if an innovation in a health care setting is “protected” by a patent or by some proprietary limit on others using the innovation, then supply could be as distorted by limitation as demand is suddenly increased, and it would be no simple matter to parse out the variance between the two.)
There is no word or phrase that economists use to describe this sort of thing; if you ask them (and I did), they will tell you that cases of this sort are rare, and have to do with very non-average elasticities of demand. So be it, but it would be nice to know the extent to which these things happen, notably in medicine as technology drives rapidly onward. So I propose to give this phenomenon a name: Cost-Pull Progress. It would be interesting to collect documentable examples from health care, but even more interesting to identify examples from other areas of the economy ravaged by cost disease. Is any empirical research going on here? I am eager to find out from anyone who knows.
Baumol Revisited. It may be that the Baumol Effect, or something like it, is more relevant to explaining cost disease than Alexander thinks. It is true, as he shows, that labor costs for teachers, doctors, nurses, and so on have not been pulled upward in real (inflation adjusted) and relative terms over time in reaction to more productive sectors. And we know the reason for the differences among sectors, which is really very simple as these things go. In virtually every industrial process it has proved possible to substitute capital for labor and make stuff both better and cheaper than in the old way. And that goes, incidentally, for monoculture agriculture, which is more capital intensive than the steel industry was just three or four decades ago.
Industries used to do this because effective capital substitution for labor made basic economic sense for efficiency and hence profit margins, but when the cost of mandatory benefits for workers began skyrocketing thanks to cost disease, the incentives for substituting capital for labor multiplied with them. Hence, in part anyway, automation mania. So cost disease has had a recursive, magnifying impact on the incentive structure of some businesses and led them to seek out new ways of balancing their fixed costs.
But alas, there are some kinds of economic activity where capital cannot be so readily substituted for labor, or where the impact of substitution is limited. Machines cannot readily replace doctors and nurses, so medical care in particular has been afflicted by cost disease in a double-whammy way: capital intensive innovation has proceeded rapidly in recent years, significantly expanding the availability of ameliorative medical procedures, but labor costs remain high as well because of the very nature of the activity. (Not that labor costs need be as high as they are and have been—there’s a difference between a fee-for-service for-profit model in health care and a Cleveland Clinic, salaried-group practice approach, for example—but that’s another topic.)
The same goes for housing, education, and infrastructure. Buildings and subways cannot build themselves, and “intelligent” machines can substitute for teachers only to a limited degree before the quality of education, as we properly understand the term, is severely affected. All the areas Alexander identifies as having been so vulnerable to cost disease fit this description: The need for labor can be curtailed only so much due to the intrinsic nature of the activity.
But what Alexander may have missed is not the average wage of workers in these fields over time, but the sheer number of jobs in these fields and hence the aggregate costs of this labor. An interesting area of research might be to establish the number of health care workers, and K-12 teachers and administrators, as a percentage of the entire labor force and of the adult population of the United States. So too with housing construction and infrastructure construction and operation, taking care to add in the OSHA and other governmental/bureaucratic personnel that have become baked into the cost structures in these domains. My guess is that these numbers have grown, perhaps significantly, over the past half century.
Here is an anecdotal illustration from K-12 education. I went to public schools in the 1950s and 1960s, and, in my experience, there was always one and only one teacher per classroom with about 25-30 students, with an occasional unpaid student teacher wandering about observing how this sort of thing is done. That ratio is now a rarity. As to management, there used to be a principal, vice principal, a head-office typist or two, and, at the high school level, a couple of guidance counselors. As non-teaching employment in schools went, this norm was pretty lean measured by today’s situation, for this is no longer the case in most schools. It would be nice to see data on the increase over the past fifty years in the average number of non-teaching jobs in K-12 schools, public and private.
The point here is that, in many cases it seems, what one worker used to do is now split down into tasks that more than one worker—sometimes three or even four workers—typically does. Seen from the perspective of the Baumol Effect, then, what we may need to do is to look not at individual salaries, but at how much it costs now to do functionally what it used to cost then. If two or three or four workers are now doing what one worker used to do, then the right way to “think Baumol” is to sum their labor costs. If we do that, the picture might look very different, although what we might end up with would be something a little different than just a more rigorous re-application of the Baumol Effect theory. Of course we need the data called for above to know what the right multiplier ought to be in various affected sectors, and it is not clear to me that we have such data.
If this turns out to be an interesting avenue of research, I think I know why at least some of this has happened. It’s the 800-pound gorilla sitting on the sofa that no one notices or wants to mention. It comes down to a technological innovation that touched off the most massive political economy change in the West (and certainly in the United States) since the Industrial Revolution: the birth control pill, and the consequent massive and rapid movement of women into the commercial work force. The pill debuted in 1960, but it wasn’t until the mid-1960s that its use became widespread. I am reasonably confident that Bureau of Labor Statistics data would illustrate what happened next in terms of the American labor force.
Far be it from me to pronounce all this a bad thing. (I do sometimes wonder how it will all look fifty or a hundred years hence, but that’s another matter.) But to just stick to the facts, the U.S. economy experienced a significant growth in labor supply just at a time when the demand for labor was less than buoyant, thanks, in part, to capital-for-labor substitution. That depressed wages for men as well as women in many economic niches, and it hurt trade unions. It’s certainly not the only thing that depressed wages and hurt trade unions, to be sure; the ur-era of offshoring U.S. manufacturing jobs began earlier than most people think.3 But it is remarkable that, when lists are compiled to account for these phenomena, the massive and by historical standards rapid entry of both well-educated and less-well educated women into the labor pool is rarely mentioned. Remember that these were “boomer” cohort women, and so there were lots of them, and remember that these cohorts of women were the first in American history to have graduated en masse from four-year colleges.
Note, too, that women who used to stay at home to shape and manage households and raise children were paid literally nothing for work worth by any measure many tens if not hundreds of billions of dollars yearly—we know this now because we know what it costs to replace that “free” labor with day care and house cleaning and restaurant expenditures and security systems and so on and on. Not that this has ever been calculated systematically to my knowledge. Were it calculated properly, we might have reason to go back and adjust the data in our national GNP accounts. We would need to add the value of this womanly “free” labor to, roughly, pre-1970 accounts, and the effect of doing that would be to rebalance the calculations of growth rates thereafter. After all, just because large amounts of non-monetized labor got transmuted into monetized labor does not mean that the economy really grew by the difference.4
Now, women joining the labor force in record numbers did not typically go into steel or other manufacturing sectors; they typically went into precisely (some of) the fields afflicted most by cost disease. It’s an open field for research to determine how the sudden boost in labor supply drove the reshaping of various areas of the economy to multiply the number of jobs “needed” to fulfill a given function. That certainly goes for women who have gone into government work at various levels—Federal, but mostly state and local (see below). We would need a Geertzian “thick description” of what has happened as well as some numbers to show us how new work protocols evolved in health care, teaching, government, and so on—in short, we’d need a sociological filter to tell us what any numbers actually mean.
Note, just to take one example, that as better-educated women poured into the monetized workforce it did not take long before women managers and HR staff directors began hiring other women—something relatively new as a mass phenomenon. No one, to my knowledge, has studied what that meant for how labor protocols developed. This research, taken all together, could constitute several dozen doctoral dissertations in sociology and labor economics, and I hope it one day does.
If my hunch is right, and if new research could quantify it at least to some extent, we might well see that some variant of the Baumol Effect can explain as least some aspects of cost disease. Someone would have to do the research, however, before we could say for certain.
Transactional Costs. Finally by way of explanatory factors for cost disease is one that inhabits several of the foregoing possibilities but that still deserves its own voice: transactional costs.
At the outset, let me try to head off possible misunderstanding of what I am about to argue. When economists talk of transactional costs, they usually mean the costs of the market infrastructure that enables transactions. They refer to theories of the firm in which managerial hierarchies can be more efficient means of transaction (especially business-to-business transactions, but others too) than continued random spot transactions. It follows that more vertically integrated companies means fewer transactions if the term is taken to mean only market transactions.
That is perfectly correct, but it’s not how I wish to use the term. Other economists, though not mainstream ones, have argued that the larger an organization is, whether governmental or not, the larger the proportion of its internal (not market) transactional costs: costs incurred just to keep the organization running. There is a biological analogy here: in small animals, like birds, the ratio of bone weight to the weight of the whole animal is exponentially smaller than it is in large animals, like elephants.5
Increased transactional costs can be worth bearing if size produces disproportional benefits, so I am not arguing that bigness alone is necessarily inefficient. But economies of scale can reach a tipping point after which further increases in size may become dysfunctional. If so, then a company that has become inefficient from becoming too large and hence perhaps too inflexible to adapt to change will be undercut, in theory, by more efficient firms. That may be one reason why the average size of economic units in the United States has fallen somewhat in recent years, perhaps because information technology-based managerial innovation in start-ups and other firms has enabled a form of internal managerial disintermediation that has disadvantaged some suddenly-too-large-to thrive firms, especially in certain fields.
This suggests, logically albeit at a different level, that we probably need to know more about the conditions under which industry consolidation as a whole, beyond individual businesses, produces net benefits for the economy and the conditions under which it produces net harm. (Institutional rentierism, I suspect, can produce benefits for the firm while at the same time producing net harm for sectors of the population and hence, possibly, for society as a whole.)
So, insofar as cost disease is concerned, consider how the average size of economic units and governmental bureaucracies in the United States has grown over the past fifty years, or past hundred years. They have grown a lot, although, again, in recent years the average size of economic units has begun to shrink back. Data here is available but elusive and, I am warned, perhaps of questionable accuracy. Maybe we can do better.
Data issues aside, perhaps cost disease is a consequence of rising internal transactional costs over time—perhaps abetted somehow (it’s not obvious how) by the need to absorb many new entrants to the labor force during the period to which Alexander’s data sets pertain—which by definition produce nothing and hence may be construed as drags on productivity increases elsewhere in the economy. This is why some economists and others suggest subtracting these kinds of transactional costs and other diseconomies from gross national product to get net national product (NNP), which would be, arguably, a more accurate way to describe productivity levels. GNP rises every time there is a car wreck and damaged vehicles get hauled off to a body shop for repair. But NNP does not rise from such events, since nothing new is produced from them.
And perhaps overlawyering, as noted above, is a special and especially insidious case of internal, not market, transactional costs generating still more transactional costs. Rule of law itself is obviously crucial to economic well being, and there should be no need to belabor the reasons. Still, every lawyerly extraction of money begets another lawyerly effort to avoid the extraction, and every new lawyer-drafted government regulation at every level begets a pestilence of lawyers hired in the private sector to “manage” the regulation to best advantage or least disadvantage.
Data on this is not easy to come by in general, but one datum is: the increase in the per capita number of lawyers in the United States over time. Per 100,000 population, the numbers look roughly like this:
- 1850 103
- 1950 141
- 1980 230
- 2000 295
- 2003 350
- 2010 400
More recent data elude me at the moment, but should not be hard to find.
Among major Western countries, the number of lawyers per capita is vastly higher in the United States than in any other country. Lawyers are expensive, and they make everything expensive as their fees are passed through the entire chain of economic transaction. Perhaps this is natural for a “Tudor” state, a state of courts and parties, as Samuel Huntington once called it. But that does not make it cheap.
So as far as a composite explanation for cost disease is concerned, the question is whether—if we sum together all of the factors adduced inside the Alexander essay and outside it as enumerated and described here—we would get close to explaining the data sets Alexander has presented? Are there possibilities still missing? The role of unions comes to mind, especially in more recent times unions of teachers, say. The role of guild-like barriers to entry that are not exactly unions but more professional associations also comes to mind, for example in the health care case, where the American Medical Association still limits the number of medical school acceptances. Are there yet others?
Of course, it is obviously one thing to establish that a given would-be cause really is a cause, and another to assign relative weights to the factors established as contributing causes. And, as already suggested, the challenge may not be merely additive, since some factors may interact with and magnify other factors. So this is a complicated challenge.
What comes out of all this, it seems to me, is a need for a multifaceted social science research project that involves economists but also sociologists and perhaps even anthropologists and psychologists. It is one worth undertaking considering the political impact of cost disease, which the Alexander essay brilliantly illustrates in its final section.
It may turn out, once the research has made strides, that the factors driving American cost disease are baked into the way we are, the way we do business, and cannot be much ameliorated. But it may be that some or many of the factors can be ameliorated by the right public policy choices. This is something we need to know, and right now don’t know. The stakes are, it seems to me, large.
1On this you’ll definitely be wanting to read Maryanne Wolf, Tales of Literacy in the 21st Century (Oxford, 2016).
2It is obvious that desirable new goods and services can stimulate increased spending, but that is not the same as costs per unit of production and so has nothing directly to do with cost disease. But, as others have pointed out, in health care, for example, it can affect costs indirectly. One way to think of medical insurance (and especially dental insurance) is not just as insurance against contingency but as partial pre-payment for medical care. All of that pre-payment cash, however, gets transmuted through markets grossly distorted by third-party payments and, under institutionalized rentier conditions, that can affect costs both of insurance and of medical care itself.
3Note that foreign subsidiaries of U.S. corporations quadrupled U.S. capital sent abroad from $25.4 billion in 1957 to $103.7 billion in 1973.
4What is especially galling about all this, on a different level, is that many women who lost their working husbands to divorce or early death ended up having very little social security earnings in the kitty to sustain them as they aged. They did not pay into the system, and their essential work on society’s behalf was never monetized in such a way as to compensate for the lack of actual withholding dollars. This was (and remains to the extent it still goes on) obscenely unfair.
5See here the pathbreaking work, still largely ignored by many economists, of Nicholas Georgescu-Roegen.