Bear with me, and I’ll make a stab at explaining a great mystery. Why have the prices of healthcare and a college education grown so astronomically for American consumers?
I’m not going to put much effort into proving that they have. For that, I recommend reading Scott Anderson’s superb, very thoughtful and well-researched blogpost on the topic. The bottom line he arrives at is that health insurance and college costs have risen roughly 10-fold, in inflation-adjusted terms, during the past several decades. (This of course does not just mean they are higher: it’s how much faster they’ve gone up than everything else.)
We often hear knee-jerk explanations of the long-term trend, and they always veer quickly into politics. For instance, malpractice lawsuits drive doctors to raise fees just to afford their insurance – the obvious solution: tort reform. Alternatively, the profit motive drives excessive medical procedures – and the obvious answer: government-mandated best-practices. In education, the blame goes to teachers’ unions or lower public funding, depending again on your politics; and either way, simple solutions are not far behind. But all these micro-explanations are undermined by a basic fact. Healthcare and education have gone up in tandem.
Anderson’s research, furthermore, focusing on the period after about 1980, shows that certain infrastructure costs have experienced an almost identical trend, of a 10-fold increase, after correcting for inflation. It seems the costs for these widely different sectors of the economy are in lock-step.
This could still be sheer coincidence. Maybe they’re each just driven by independent cost dynamics that all just happen to work out in a similar pattern. Except that the prices we’ve ended up with after these multidecade trends are quite unusual in world practice. Healthcare and education are both offered in the US at a multiple of what they cost in other First World countries; the same is apparently true for public transport. What an amazing coincidence. Are the independent drivers of these separate cost curves also all independently unique to the United States?
What’s called for is a macro explanation, and I’d like to offer one here. It will require a detour into actual economics, but I’ll try to keep the model descriptions brief. (Links to the original articles, with diagrammatic and algebraic modeling, are provided.) That detour, in turn, will give us some useful tools for describing what’s going on — and may even shed light on certain other peculiarities emergent in the US economy over recent decades.
The traditional explanation
Let me begin with the explanation most commonly cited by economists. This is something called the “Baumol effect.” What’s rather wonderful about this idea, first explored in 1967 by an economist named (as you might guess) Baumol, is that it does account for the tendency of prices in certain sectors of an economy to increase much faster than in other sectors; and moreover, for the most part, it correctly identifies which ones.
The basic Baumol concept is that imbalances in productivity gains are driving the whole process. Here’s how it works:
Say you have an economy consisting of only two sectors, call them 1 and 2, such that Sector 1 includes all those fields of activity that are more-or-less defined by the amount of labor they contain (I’ll explain this in a moment), while Sector 2 includes all activities subject to automation. Education and medicine would belong to Sector 1, because the nature and quality of the service depends crucially on how much human attention you get. You can add devices and rationalize to some extent, but you can’t reduce the human factor indefinitely. Sector 2 is the opposite. It unites those areas where robotics and rationalization can bring unit labor inputs down almost without limit, often while dramatically increasing output. Now, Baumol pointed out that the labor productivity of Sector 1 is effectively frozen, while in Sector 2 it can enjoy geometric growth. If the wages of the two sectors were directly proportional to productivity, which he felt they probably are, then employees in Sector 2 would end up making a lot more money than in Sector 1. Pretty soon, in fact, Sector 1 would just be emptied out of employees altogether, because they would all prefer to go make the higher wages paid elsewhere. Or else Sector 1 could offer higher wages too, in order to compete with Sector 2 and hang onto its employees; but to do that without being able to increase productivity, Sector 1 would have to charge higher prices.
There’s a beautiful logic to this. And over time, a process like this would lead inevitably to some of the phenomena that we do in fact observe today. Thus we do see labor-intensive goods and services priced higher and higher. A twist on this is that as their prices increase, demand for some of these items shrinks away until they become luxuries unaffordable to anyone but the elite. Baumol gives examples, citing handmade furniture and ceramics as well as theater tickets; and in fact it’s true that low-productivity goods such as custom-made items and live performances are priced beyond reach for many today. Their role in the economy has been substituted almost entirely by machine-made goods and recordings. Yet at the same time, in Baumol’s model, there are certain other low-productivity services that keep raising prices for those same reasons, yet that don’t run into the demand barrier: namely, those services that people generally consider essential, such as medicine, for instance, or perhaps education too. These are low-productivity goods as we defined the term earlier (i.e., Sector 1 goods), but demand for them is highly price-inelastic. Their prices can keep growing ineluctably without demand shrinking away.
Yet as explanatory as this approach is, it is clearly not the whole story, or even most of it. In fact there’s something charmingly naïve about Baumol’s theories, setting them apart as a product of a bygone time. In the 1960s automation could still seem simultaneously the driver of human progress and yet also the cause of certain regrettable consequences that were nobody’s fault. The times since then have made us more cynical.
Let’s start with the fact that wages in Sector 2, the one subject to automation, have not kept pace with labor productivity. Not anywhere close. As a result, there’s been no need for Sector 1 to struggle, keeping pace with the higher and higher wages of factory employees. Something like the Baumol effect may in fact apply to certain artisans trying to eke out the cost of living from their handicrafts; but it seems a little absurd to stretch the case to doctors, many of whom are millionaires. The underlying truth is that, yes, productivity has been increasing, but the real wages of employees have been going nowhere. In fact, real wages in the US peaked some 45 years ago — i.e., before the truly astronomical price developments we’re discussing here. Moreover, the companies in the US manufacturing sector, far from driving wage-growth through their success, are under pressure, holding wages down, laying off staff, or else gradually disappearing altogether.
The reason the Baumol discussion is ultimately unsatisfactory, in my view, is that there’s only so much we can understand without reference to the foreign sector. And this brings me to another model, more commonly used to analyze the problems of developing countries; but I think with a few tweaks it has great explanatory power for the US as well. Dating back to 1982, this one is the brainchild of a couple of authors named Cordon and Neary, and it was first invented to explain a curious syndrome called “Dutch disease.”
What is Dutch disease?
Dutch disease is an imbalance that has been observed in some countries, such that booming exports can cause them to suffer widespread deindustrialization and at the same time a skyrocketing cost of services. Let me introduce the Cordon & Neary model as briefly as I can. I’ll start by describing it in its simplest classic form, and then go on to show how it may describe what’s happening in the US.
Assume an economy producing only three outputs. Two of these are goods that are freely exported and imported; they are called “traded goods,” because they’re traded on world markets, at prices determined by world-market trends rather than by anything going on inside our hypothetical economy. For the sake of convenience, we’ll designate one of these two as “Oil,” and the other as “Manufactures.” (But do note that the model allows other definitions.) And the third sector of the economy consists of “non-traded goods.” These are different by virtue of being neither exported nor imported. We’re going to designate them as “Services” because services, generally speaking, such as medical care and education, are priced and consumed inside the country where they are offered, not on world markets. That said, this “non-traded” sector could include other essentially domestic products such as fixed infrastructure.
The question in this scenario is: what happens if there’s an export boom in one of the traded sectors, for instance because of a major increase in the price of Oil on world markets. (Assume that this economy is operating at full capacity.)
Well, according to the Cordon & Neary model, one outcome is a little reminiscent of the Baumol dynamic we were just discussing. Namely, the booming Oil sector starts paying more wages in order to expand by attracting more workers. This is the so-called “resource effect,” because it tends to draw resources away from the other two sectors of the economy. Those other sectors then have to raise wages in order to hang onto their personnel; and in order to do this, they have to raise prices. Thus the price of Services goes up, and (depending again on price-elasticity) this may result in lower demand for them. For Manufactures though, the equation is different. Any increase in prices simply results in worsened competitiveness against imports and a direct loss of market share. On balance, the “resource effect” of the Oil boom is twofold: the price of Services goes up, and there is a deindustrializing pressure on Manufactures.
The other effect of the export boom, called the “spending effect,” operates as a result of economic growth. The increased income to the Oil sector causes economic activity in general to heat up, i.e., there is increasing demand for all goods and services. For Services, facing little or no foreign competition, the increased demand is equivalent to a higher volume of sales at a higher price point. For Manufactures however, the heightened demand is leaked away in the form of additional imports, while the cost pressures of a heating economy simply add to the momentum of deindustrialization.
Currency effects and real appreciation
It’s worth looking at the currency perspective on what we’ve just described.
First even if we ignore monetary effects and nominal exchange rates (as Cordon & Neary did in their 1982 paper), the dynamics still equate to a change in international terms of trade. Namely, the local factors of production become more expensive, as measured by their world-market buying power, than they were before the export boom. This is clearly true for wages, for instance, which go up nominally while the price of imported Manufactures stays fixed. And as a further result, as we’ve seen, local Services become more expensive relative to imported goods as well. This general repricing of local inputs relative to foreign goods, this real appreciation of the local currency, is in fact the underlying reason for the deindustrializing effects faced by the lagging traded-goods sector (“Manufactures”). But that’s not the end of the story, because these dynamics tend to be strengthened further through explicit nominal and monetary effects (not part of the original model). Let me show this briefly before finally turning to the United States.
It should be clear that our hypothetical export boom would tend to cause not just real but also nominal appreciation of the local currency. As the exporters bring in additional foreign exchange, in the absence of monetary accommodation, a growing amount of foreign earnings will be chasing a fixed amount of local currency, thus driving the exchange rate up directly. Moreover, the export boom may stimulate other important inflows of foreign currency as well. For instance, foreign banks may want to invest in the Oil sector, and their inflows will further drive up the local exchange rate.
The alternative is for the local Central Bank to intervene in currency markets, creating more local currency so that a growing money supply meets the growing export proceeds, in an attempt to hold the nominal exchange rate more-or-less steady. This happens very commonly in such situations, and it can help the local economy remain more globally competitive. Importantly, an expansive monetary policy of this sort then further stimulates the economy, fueling GDP growth and demand even beyond what they would be from just the export boom alone. As a result, the monetary intervention strengthens the “spending effect” factors discussed above, with the result that the local currency’s real appreciation is accelerated. Moreover, this loose monetary policy can lead to strong inflationary pressures. When this is combined with a fixed exchange rate, a local currency that is losing its domestic buying power retains its nominal forex value.
What if, however, monetary policy is tightened as inflation arises? When that occurs, the very fact of higher interest rates will tend to attract additional foreign currency inflows. Thus as inflationary pressures are curbed, counteracting the “spending effect” in the economy, which as we saw was leading to real appreciation, the result is to heighten international demand for the currency, with the result that nominal appreciation will bounce back. Policymakers can attempt to thread this needle, and there are other policies they can try (based on so-called “sterilization” of the money supply); but eventually a sustained major foreign currency inflow will lead, one way or the other, to a change in the country’s international terms of trade, most often accompanied by nominal exchange-rate appreciation.
How does any of this apply to the United States?
These scenarios are not just hypothetical. Dutch disease has been identified in a number of famous cases around the world, as a glance at Wikipedia will show. The syndrome is in fact rather common, particularly in developing economies with a booming raw material export sector. But how on earth could this be relevant to the United States? We’re not a developing country, and we are not experiencing an export boom, right?
No, but we are nevertheless, and have been for a very long time, the recipients of a massive inflow of foreign funds. This is similar in some of its effects to the earnings that would be generated by a single, very powerful export sector. The difference is that our inflows of foreign exchange consist of foreign investment. And the most powerful engine of the process is not exports but our fiscal deficit, which sucks in investment from around the world. The massive foreign funding began roughly with the presidency of Ronald Reagan, when the US began running so-called “dual deficits:” that is, large federal budget deficits, coinciding with large foreign trade deficits.
The average of our federal budget deficits over the entire 1981-2017 period, that is, beginning with Reagan, has been about $340 billion per year: a total of some $12.5 trillion during that entire time. This corresponds rather closely to our foreign trade deficit over the same period, which has averaged $310 billion per year and has totaled $11.5 trillion. Now, it’s not actually true that the entire cumulative federal budget deficit during that 37 years has been financed directly by foreign lending to the US government. In fact, only about half of it has. Foreign holdings of US federal debt are currently around $6 trillion. The remainder of our federal debt must have been purchased by domestic investors. But it is nevertheless fair to say that our trade deficit had to be financed by foreigners somehow. Thus very large, sustained inflows of foreign credit and other funding did take place, roughly half of which was direct lending to the US government. The other half must consist of other credits, investments and transfers. The point here though is not to identify all the segments of that foreign investment specifically, just to show that it exists and that it corresponds roughly in size to the US federal deficit, which therefore seems to be one of its primary drivers.
It is no coincidence that this long period, characterized by consistent, massive inflows of foreign funding, has been the same as the long period we were discussing at the outset of this article, during which the costs of key (“non-traded”) services have been accelerating out of control. This is precisely what the Dutch disease model would predict.
But if we do look more closely at the idea of Dutch disease in the US experience, and if there is no export boom in the narrow sense, then where should we look for the effects attributed in the model to the booming traded-goods sector, i.e. the goods that we called “Oil?” The answer will require some discussion, but let’s start by having a look, first of all, at the financial sector, which has handled all those flows of foreign funding, earning enormous fees and commissions in the process.
The US financial industry has largely come into its own from near the beginning of the Reagan presidency, when those dual deficits came into play. Just as a very rough and ready indicator, as of 1982 the S&P 500 Index was about where it had been 28 years earlier, in 1954; and in fact it had reached that same level even earlier, in 1937. But a relatively sustained revaluation seems to have begun during the early years of the Reagan administration, such that the broad S&P index has gone up more than tenfold since then. And not surprisingly, finance has been on a wild ride for much of that time.
Before the 1980s, banking was dominated by neighborhood S&Ls and state-level commercial banks. It was still possible to live a full life without a credit card. Few middle-class families owned shares in companies; and for most people the stock market was an exotic topic, outside their normal frame of reference. Since then, the financial industry has exploded upon us with a plethora of retirement and investment and liquidity-management products, speculative packagings and repackagings of risk, a deregulation of highly-leveraged, rapidly-growing national and now global investment banks, and so on. The institutional development, the widening range of products and the sheer turnover growth have all been simply incredible. If US finance has not been a “booming” sector over that 1981-2017 period, it’s hard to imagine what one would look like. And as the Dutch disease model would predict, the sector has been one of the leaders in terms of, not wages exactly, but reward for labor. There’s probably no need to mention hedge funds and Wall Street bonuses here.
This was not a narrow boom, and the rewards radiated outward in complex patterns that are beyond my scope. Clearly the legal profession got a boost from the wholesale re-thinking of institutions, there were major technological revolutions as well, and the disintermediation of financial services was both a stimulus for and a result of the growth of online businesses and software products. I don’t mean to oversimplify the immense US economy to the point of attributing all growth to a single factor. It’s enough that we recognize that factor’s existence among others. In any event, there has been the $310 billion per year net inflow described above, channeled through the financial sector, and the financial sector has been booming as one would expect.
Meanwhile, again true to the Dutch disease model, a significant part of the US manufacturing sector has been subject to deindustrialization pressures. In what appears very similar to the so-called “resource effect,” US-based companies have consistently found that their local factors of production – labor, rental and others – are simply getting too expensive to allow effective competition against imports. In response, there are plant closures, outsourcing and relocation of production overseas.
Now, it may be worth mentioning a nuance here. I’ll confess that the Dutch disease model is meant to apply to what’s called “small, open economies,” and of course the US is anything but that. What’s meant though is merely that the prices for imported goods in the model are considered exogenous, for all practical purposes fixed. That is, import prices are not a function of the importing country’s own demand. That’s not true for the US, strictly speaking. However, at the risk of obvious oversimplification, I would argue that it is true enough for our current discussion. Factors such as the long-term poverty of many of the exporting countries, as well as the desire by corporate and various national interests to gain or hold onto market share, have kept US import prices relatively low. In terms of their competitive effects on US producers, we can say that the price dynamics are reasonably close to what the Dutch disease model assumes, in that they clearly do confront US producers with outside price constraints.
Moreover, it’s interesting to consider the impact of another fact: that the range of imported goods has been widening. This is a result of tech transfer and production-site decisions by US companies, as mentioned, but also more generally it’s just a feature of the increasing economic development of many Third World US trading partners. In terms of our Dutch disease model, what it means is that the sphere of the US’s lagging traded-goods sector (“Manufactures”) has been widening as well. We looked earlier at real appreciation, saying that it consists of a repricing of local inputs relative to foreign goods. My point now is that this also coincides with a terms of trade shift within the United States itself, between those sectors that are driving local input costs higher and those whose prices tend to be constrained by import competition. What’s worth considering is this: as the range of imports widens, the proportion of these two sectors changes. The import-constrained share is growing, and the independent-pricing share is shrinking. The real appreciation of the dollar thus translates into purchasing-power gains for a smaller and smaller share of the US population as the sphere of deindustrialization expands.
The US “non-traded sector”
As shown, we have several of the features of Dutch disease: our foreign inflows, our booming sector(s), our real revaluation effects, and our deindustrialization. Should it really be a surprise that the remaining feature, a skyrocketing cost of the “non-traded sector,” is happening as well?
In an economy suffering from Dutch disease, the model shows us that the non-traded sector is the main avenue available for expressing price pressures. Looking specifically at the US economy, cost-growth in terms of nominal inflation is of course held in check through a proactive monetary policy; but bear in mind that the consumer basket used to measure CPI includes imports. As the sphere of deindustrialization expands, nominal inflation will tend to be increasingly restrained by this factor — and price developments in the non-traded sector will uncouple from the inflation rate. Thus the prices of the services we’re wondering about (medicine, education, etc) will outpace inflation; and moreover, their price gains can be expected to accelerate relative to inflation as deindustrialization proceeds.
Let’s return to the “resource effect” concept discussed earlier. As we know, with foreign revenues pouring into the US, the booming sectors of the economy are willing and able to pay a premium for resources such as labor (but also other resources such as credit, office space, etc – labor is a shorthand for all the local production factors). This upward pressure on local costs tends to force other sectors of the economy such as medicine, education, etc, to pay more too.
Does this suggest that a doctor will become an investment banker if he isn’t paid enough? No, but the costs of running his practice will be influenced by those willing to pay more for staff, office space and other resources. Moreover, the lifestyle a doctor expects will inevitably be influenced as well, by comparison with the booming sectors of the economy; and over time so will the career choices of young people. With operating costs and expectations both on the rise, and with price-inelastic demand imposing only very limited constraints, healthcare prices go up.
There’s another perspective on this as well. Namely, we’ve mentioned a few times that the “non-traded” parts of the economy are not all equal. Certain services, we’ve said, have a very low price elasticity: people in need will pay nearly anything for them. Other services though cannot engage in a bidding war for labor or other resources, financed through price increases, because their own demand can’t sustain it. As an example, consider fast-food restaurants or household maid services. Demand for them is surely not going to allow sustained, indefinite price growth, so in our model the resource effect should mean that either demand for them shrinks away (if they try to raise prices) or else they lose commercial viability (if they don’t). However, in practice we see that they are thriving. Why is that?
The answer is quite simply that, unlike in the simple Dutch disease model, we don’t have full employment or a perfect mobility of labor. Thus some of these services, despite cost pressures, are in fact unable to raise prices substantially due to demand constraints; yet they can overcome this to a large extent by simply holding the line on wages. Due to persistent unemployment and other factors, their employees are unable to migrate into higher-paying activities; thus the cost pressure on such services is expressed as immiseration of their workforce, which may very well coincide with thriving businesses. It’s not that the sector as a whole isn’t facing the “resource effect;” it’s just that much of the pressure is successfully shifted onto labor. Note that the Dutch disease model would expect this immiseration effect to be least-pronounced with respect to purchasing-power in terms of imported goods; and most-pronounced with regard to price-inelastic services such as medical care. This helps explain the objectively rather strange price structure we observe in practice: a fast-food worker might be able to afford a 60-inch ultra-HD television, but not a doctor or a college education.
Thus the Dutch disease model, with a few tweaks, predicts an all-too familiar pattern: widely available and cheap consumer goods; a highly prosperous segment of the economy; a sustained growth, above the rate of inflation, in the price of inelastic goods and services such as healthcare and education; deindustrializing pressures throughout much of the manufacturing sector; and a segment of the economy that is priced out by all this and left behind.
If this dynamic appears to tell us what is happening, of course, the related question becomes, what to do about it. Unfortunately, the Dutch disease model does not contain an answer. It appears to me that there might be numerous policy initiatives that could be helpful. Regulation of certain sectors known for their price-inelastic demand functions could mitigate some of the impact, particularly for the poorest members of society. Seeking more balanced fiscal and foreign trade policies might help address the issue at its root, relieving some of the pressures of cost growth and real appreciation. In case we ultimately decide that some level of deindustrialization is inevitable, or even desirable, perhaps we could undertake measures such as retraining to manage the social consequences. In any case, one thing seems certain: a first step would be to understand the underlying commonality of the trends we have been experiencing for so long.