Consumption in the U.S., per capita, measures about 50 percent higher than in the European Union. American individuals command more resources than people in countries such as Norway or Luxembourg, which have higher per capita GDP. The same American consumption advantage is evident if you look at dwelling space per person or the number of appliances in a typical home.
He says to look at total U.S. health care spending relative to the size of our economy in this context. If you put GDP in the denominator, we look like an outlier. But if you put consumer spending in the denominator, it looks fairly normal.
And yes, I too have cited the Random Critical Analysis piece.
Funny how nobody looked at this before. I vaguely understood it when I would point out that of course spend more on healthcare and that our government already pays in dollars per capita what others pay.
The obvious initial implication is that government should pay for (or rather, should not pay more than) basic coverage, since the excess spending is partly voluntary.
EVERY left wing talking point to push Single Prayer (TM, copyright, Andrew’) that Republicans have accepted is wrong. I showed this daily for 16 months at MR.
Thanks for the link.
It’s worth pointing out that it isn’t just health care that correlates much better with AIC. Individually and collectively (with dimension reduction techniques like PCA) the other major consumption categories (pertinent headings under COICOP & COFOG) correlate much better with AIC than GDP in cross sectional (with volumes/sector specific PPPs and with GDP PPP adjustments) and in time series analysis with the available OECD data in various configurations (expenditure and expenditure shares). Put differently, as overall consumption (AIC) rises there are some pretty strong patterns in how countries choose to allocate their consumption expenditures. The share of expenditures allocated towards different consumption categories, not just health, shifts markedly and it does so in a way that is rather well related to AIC (not so much GDP). The US also appears to be very much on trend in its consumption composition as a function of AIC in these different specifications.
I find this mostly unsurprising (it’s what I would intuitively expect), but it is a useful exercise if people think this is only some sort of statistical fluke that is limited to health expenditures or that AIC is merely some sort of arbitrary construct. So far as I am concerned whatever combination of factors (e.g., productivity, different economic structure, economies of scale/population size, cultural differences in propensity to consume, fiscal policy, etc) come together to produce a certain level of AIC (or equivalent) that’s still the relevant margin we need to judge against (or at least much closer to it than GDP) if we are trying to ascertain the efficacy of different health systems vis-a-vis cost containment and the tradeoffs involved.
My question is what is the relative contribution of the real and nominal factors in high American health care consumption.
That is, how much of higher US consumption is due to higher prices for the same level of goods and services on the one hand, and how much is due to more goods and services being consumed per capita, for similarly situated patients. And are those relative contributions consistent across patient types, or are there significant differences for certain consumer categories (e.g. end of life care).
The answer to the above question has a major influence on how one ought to interpret the result.
I addressed the role of relative price differences in my latest blog post. The best available evidence suggests US prices are generally consistent with what we would expect from a high income country and that our issue, like other countries, largely comes down to high volume (OECD estimates suggest the real volume of health consumption is something like twice the OECD average) and that US growth in health expenditures relative to income over the past few decades (at least) is almost entirely explained by high and rising volume.
I think of all the thousands of dollars we spend on tests, etc. that are generally not worth the cost and that we would never spend in a system more regulated by our personal spending decisions.
If this analysis is true, then that means we wouldn’t necessarily spend less on health care in a more effective system. We would still tend to spend a similar amount. But, we would spend it on more useful services. Imagine having 40% more real healthcare spending. That would be amazing.
This is what I have found in housing, also. For an item so large, income effects dominate. Whether we were living in palaces or sod houses, we tend to spend about 25% of incomes on housing, over the long term. This means that real housing is essentially free – and I think data over time bears this out. If we create more housing, we still spend about 25% of income on housing, but we just get more real housing for it. It’s free.
How is that? Resources were used to create the new housing. It is true because, while resources are used to create new housing, the addition of new housing means that existing real estate owners claim lower income. Economic rents flowing to existing housing capital decline. We can see this balance in the extreme in the urban cities where a long term dearth of new housing stock has led to extreme economic rents claimed by real estate owners.
On the other hand, in housing, targeted tax subsidies do seem to tend to cause owners to spend until their housing expenditures, minus the tax benefits, still add up to that 25% spending level. The same should be true in medicine. So, to the extent that, say, employer tax deductions for health benefits subsidize extra health care spending, surely this induces more nominal spending as a portion of incomes.
Just don’t confuse consumption spending with consumption spending by individuals as a significant portion (about 20%) is government spending on the behalf of individuals.
But the expected life span doesn’t increase in the same way as dwelling space per person or the number of appliances in a typical home. What is the metric used to indicate the expenditure increased value.
That’s the hard part.
My ACL repair didn’t extend my lifespan but it allowed me to continue training. My ortho said at my age you only do an ACL repair if you want to do something that torques the knee. So arguably if I switched to something like biking I wouldn’t need an ACL repair.
What’s the value of those types of treatments? Should life quality treatments be self-pay? If so who decides what constitutes a life quality treatment? Should low income individuals get subsidies to afford them? If so, who decides how much the subsidy is? Are providers required to accept that subsidy amount?
Multiply that by the 1000 marginally overlapping markets that are cumulatively described ‘health care’
I suspect part of it is cost disease. You get a hyper-over-qualified American doctor to spend 5 minutes on your problem.
And second of all, which should habe neen first of sll, I just did it too. The big lie becomes ingrained in the brain.
The left wing talking point that we don’t have higher life expectancy and that our healthcare doesn’t deliver is simply false.
Forbes: The Myth of Americans’ Poor Life Expectancy
Of course spending twice as many dollars doesn’t get us twice the life expectancy, but not only would that be silly, we can’t buy as many American made hats as we could if we exchanged dollars for pesos and bought Sombreros across the border. Stuff is more expensive here.
We probably buy more healthcare because we do get something for it.
But we are well into the diminishing returns zone.
But it is your freaking life and that of your loved ones you stingy cheapskate!
Love your work in general, but let me illustrate two societies in two different situations to shed light on one of my issues with this.
Scenario 1:
Society A
Income = 160
Health Expenditure = 40
Housing Expenditure = 40
Education Expenditure = 40
Goods Consumption = 40
Society B
Income = 120
Health Expenditure = 30
Housing Expenditure = 30
Education Expenditure = 30
Goods Consumption = 30
Let’s say that healthcare, housing, and education represent zero sum positional goods. We’ll ignore additional square feet in the USA, we are talking about SF real estate here. The only goods that have a marginal benefit to societal utility is the last “Goods Consumption” category.
It’s not clear what additional benefit society A gets form the additional spending. In theory its “33% richer”, but only 10 of the additional 40 units of “wealth” are going into utility increasing economic activity.
I don’t think we can really pull the “its what people are choosing” card. We are talking about a notoriously non-transparent third party payer system. It’s hard to believe anyone whose received extensive healthcare in the US thinks rational economic trade offs are being calculated by utilizers of healthcare (and its not their fault).
I’d be much more impressed by:
Society C:
Income = 160
Health Expenditure = 30
Housing Expenditure = 30
Education Expenditure = 30
Goods Consumption = 70
If the top three categories were yielding the same results. If they also minimized the economic dislocation of the US system (medical bankruptcy) that would be a big plus.
Then of course there is a issue of debt. One thing I don’t understand in your model is that savings and assets aren’t considered. So Norway being able to afford both good healthcare and high savings means they have worse quality of life?
Let’s look at this again with debt.
Scenario 2:
Society D
Income = 160
Health Expenditure = 40
Housing Expenditure = 40
Education Expenditure = 40
Goods Consumption = 40
Savings = 0
Society E
Income = 160
Health Expenditure = 30
Housing Expenditure = 30
Education Expenditure = 30
Goods Consumption = 40
Savings = 30
Society E is obviously a lot better off then Society D, even though non zero sum consumption levels are the same. Society E is building the capital to produce future income and provide a store of capital for crises. Society D isn’t saving enough because its spending it all on zero sum consumption. If I’m interpreting AIC right aren’t you ignoring savings?
I agree with your analysis.
The problem is that the reason they charge $15 for the aspirin is they don’t know how to do otherwise.
We could have politicians say they have to lower the price on the aspirin, but they’ll shift it somewhere else.
Let’s say that healthcare, housing, and education represent zero sum positional goods.
Wow. Just wow.
Let’s say that healthcare, housing, and education represent zero sum positional goods.
Wow. Just wow.
How far off do you think it is?
I’m not going to agree with him on housing, but on education my first estimate would be 100%.
I am splitting up this response into multiple comments because I have had poor luck with automatic moderation systems in the past with link heavy comments in wordpress in the past and I think splitting should work.
#1: You are kind of missing the thrust of my blog post. My main point is that regardless of whether this spending is actually optimal, US health expenditures are quite well explained by AIC and there’s not much empirical evidence to suggest that the different health systems found in other countries actually cause them to spend substantially less (or that their expenditures go markedly further conditional on income levels). Put differently, even if one concedes it’s necessarily suboptimal it’s not a solved problem and the other countries generally behave quite similarly on the relevant margin. Trying to mimic the health systems found in other countries is very unlikely to deliver notably less expenditure or better ‘outcomes’ as the cause of our extra funding (and worse outcomes) largely relates to factors that have little to do with unique features found in our health system.
Thanks for your reply. I’m sorry if this response seems short. My wife is getting induced on Monday and I doubt I’ll have much time to continue this exchange.
When I take a look at this graph:
https://randomcriticalanalysis.files.wordpress.com/2016/09/rcafdm_54_who_and_worldbank_nhe_pct_aic_by_aic.png
I see the USA at ~23% of AIC. Then we have Germany, UK, Canada, France, and Australia (just choosing some large OECD pack countries) clocking in at say 13-16%. That’s using your preferred metric of AIC.
How are these in alignment? One number is higher the the others, and that is even using AIC as a denominator to make the US look better. The explanation you seem to offer is that “healthcare is a superior good”, and that as countries get richer their AIC% (keep in mind, I’m still talking %) goes up.
That seems speculative though. Most of your dots in-between the USA and those countries are tiny little banking/oil states (Luxembourg, Norway, Switzerland). Ones where you use AIC because its way lower then GDP. I’d note that these are countries full of very wealthy people (who can afford expensive healthcare) and that have high savings rates compared to the USA (which I think is relevant to societal well being).
I’m not sure that’s enough to conclude that once Australia is as rich as the USA it will have the same level of AIC% spending on healthcare.
Maybe they will be able to maintain their level of AIC% relative to the USA even as AIC per capita increases. I guess we won’t know till it happens.
Imagine a world where the USA maintained an OECD level of AIC% spending that was significantly lower. That 8% or so could be used for a higher savings rate, extra material goods, etc. Things that really improve quality of life.
I can buy that my new car with the blindspot sensor is a superior good to an older model of car, but then again I can measure that in how my insurance rate went down because drivers have fewer crashes.
I question healthcare as a superior good. We don’t see much actual benefit to this higher spending whether measured in absolute or relative terms. I’m not just talking life expectancy, outside of a few areas there doesn’t seem to be much measurable quality of life improvement. As a lifetime consumer of much healthcare and working in the industry it feels like a lot of wasteful spending is going on because its all other peoples money.
Example: Was Crestor really better then Lipitor? Back when it was a brand we were paying over $200 for Crestor and $4 for Atorvastatin, but health plans were falling over each other to offer Crestor. And yet whenever we could pass the cost onto the consumer all of a sudden people didn’t need Crestor anymore, all that statins are the same it turns out! The doctor that prescribed me Crestor did say that the tits on the drug rep they sent over were pretty good though.
I didn’t feel like I got worse care while living in Japan, its not like there was something I needed in America that I couldn’t get there, even though they spend much less. Not having to deal with the byzantine American medical billing system was also a giant quality of life improvement.
Maybe if they ever get as rich as us their AIC% will go up, but its just a conjecture at this point based on basically ourselves and a few tiny countries where AIC diverges wildly from GDP.
> My wife is getting induced on Monday and I doubt I’ll have much time to continue this exchange.
Congrats and good luck.
> I question healthcare as a superior good. We don’t see much actual benefit to this higher spending whether measured in absolute or relative terms.
When I say it’s a superior good I mean this purely descriptively to point out that the (raw) elasticity of NHE with respect to GDP (or AIC, disposable income, etc) is clearly much greater than one. Whether countries ought to spend like this is somewhat of a different question (depends on values, assumptions, etc), but my point is AIC is clearly a stronger predictor of NHE than GDP and largely explains US, i.e., it suggests the cause of high US NHE lies largely outside of the idiosyncratic differences of our health system.
> When I take a look at this graph
That WorldBank/ICP derived graph implies elasticity much greater than one, but we don’t need to rely on this alone. As I documented in my blog post at some length, we find the same thing with OECD panel data (more reliable statistics plus temporal dimension). I think this animated GIF is particularly instructive because you can clearly see NHE rising faster than AIC for ~all countries that have experienced significant growth in AIC (which tends to argue against endogenous differences being a significant explanation the slope found in cross-sectional analysis) and that most countries seem to be following a very similar long term trend. Countries that today enjoy levels of consumption comparable to what the US did 1-3 decades earlier allocate quite similar shares of AIC towards NHE (more in many cases) and there’s no evidence of this slope flattening as they get richer (if anything quite the opposite!)
> Most of your dots in-between the USA and those countries are tiny little banking/oil states (Luxembourg, Norway, Switzerland)
That’s part of my point though! US consumption levels are so much further ahead than the rest of the OECD in recent years, even if you subtract things like health and education out completely, that few are even in the same ballpark (same goes for adjusted household disposable income), ergo comparing NHE per capita and the like isn’t really the indictment of the US health system that many think it is. Although we have essentially no truly comparable peers, we can:
(1) compare historical trends for US and other countries [as mentioned above] Despite the fact that the presumed flaws of US haven’t much changed, US NHE wasn’t always particularly exceptional. This corresponds quite well to the increase growth in household material living conditions.
(2) use OECD panel data to train a model to predict NHE for non-US OECD countries and compare fitted values to observations for the US. I did this here with a non-linear specification and found the US was even closer to fitted values (note: time/technology effects appear to be very modest as real AIC explains almost all growth–technological adoption is the primary means by which countries increase expenditures, but same technology gets cheaper with time and without increased real AIC countries tend not to increase expenditures by much more than you would expect)
> Ones where you use AIC because its way lower then GDP.
I use AIC and find it to be a credible explanation of NHE because I have found it to be a much stronger predictor at multiple levels of analysis. Once you know AIC (or adjusted household disposable income) the remaining components of GDP offer virtually no additional incremental validity. Even if you crudely subtract NHE out of these measures first the conclusions are much the same.
Incidentally, when comparing against the likes of Norway etc with GDP the issue is not so much that US AIC share of GDP is exceptionally high (it’s really pretty close to OECD average), but that these small oil/banking countries have exceptionally low AIC shares of GDP and this actively confounds the GDP trend (especially w/ range restriction in cross-sectional analysis on the handful of richer OECD countries).
> I’d note that these are countries full of very wealthy people
And yet US households have much higher financial wealth than all but maybe Switzerland and has higher household savings rates as compared to household disposable income than the EU as a whole (and DI is much higher in the US). I mean I’m not saying cultural differences in propensity to consume play no role in AIC whatsoever, but it’s just not an especially large one.
Countries differ for quite substantially other reasons, in terms of economic structure, size/scale, threats, etc, that makes naive comparisons of GDP as measure of the material welfare of domestic households quite misleading. I do not believe it is a accident that essentially all of export dependent countries in oil, banking, etc diverge systematically from trend and that AIC explains ~all of these apparent outliers quite well.
In any event, if you really want to learn more about this topic and why AIC and adjusted household disposable income are likely to also be theoretically superior predictors than GDP I highly recommend you read this long paper by Joseph Stiglitz, Amartya Sen, and other notable economists on GDP and its alternatives (they specifically recommend these measures, albeit under a slightly different name for AIC)
Only heave time for this:
Countries that today enjoy levels of consumption comparable to what the US did 1-3 decades earlier allocate quite similar shares of AIC towards NHE (more in many cases) and there’s no evidence of this slope flattening as they get richer (if anything quite the opposite!)
1-3 decades ago is a long time. Medical technology and a number of other factors are different.
I think its fundamentally speculative to assume that per capita AIC drives AIC% of healthcare expenditure. You’ve only got one large country OECD data point (the USA) and your assuming that per capita income is the driver.
The driver(s) could just as easily be medical technology. Or aging. Or several at once. Per capita income rises over time everywhere. You could pair it up with just about any metric that also rises over time. There isn’t enough evidence here to reach your conclusion.
To really get definitive you would need another large OECD country with the same per capita income in the same time period.
Instead, I think this speculation could go on forever. Each year AIC% will rise in all these countries, but our AIC% will continue to remain significantely above theirs. Each year their AIC per capita will be lower then ours, because its not like that is going to change anytime soon.
Put another way, can you think of a real life test that would answer this theory definitively and that we are likely to actually observe?
All we really know today is that given the current global economy and current availability of medical technology these other countries are providing medical care of identical effectiveness both at a lower per capita rate and a lower % of AIC. They also do it while eliminating a billing system that causes infinite frustration and medical bankruptcy.
asdf: I am responding up thread since we’ve hit the thread limit.
> 1-3 decades ago is a long time. Medical technology and a number of other factors are different.
I mean 1-3 decades worth of growth in AIC, i.e., approximately projecting each countries’ growth out how long will it take them to reach the material living conditions the US enjoys currently (~2017). In some OECD/affiliated countries it’s 40+ years and in a small handful of others it’s *maybe* as little as ten (note: Switzerland and Luxembourg respectively “only” spent around 20% and 28% less than the US in 2015).
> To really get definitive you would need another large OECD country with the same per capita income in the same time period.
I disagree, RE: speculativeness. Obviously we can only make quantitive educated guesses until this happens, but the same very much applies to arguments against the US system (the vast amounts of money we’d presumably save if only we aped the modal OECD country) and reliable data are much more consistent with differences in AIC.
For example:
1) Changes in US health spending track well with changes in disposable income in long term. When US disposable income falls NHE falls with it and when disposable income rises NHE rises (albeit with a modest lag factor of 2-3 years) [note: CMS uses disposable income in their models over GDP].
2) Similar patterns are seen elsewhere, vis-a-vis disposable income/related and NHE. In the last financial crisis those countries that experienced particularly large negative shocks in their disposable incomes (e.g., greece, italy, etc) saw their NHE plummet (and I’d expect some “stickiness”)
3) Up until the mid 70s/early 80s US NHE was not exceptional when measured in per capita terms or even % of GDP. The apparent divergence in NHE coincidences neatly with the divergence of AIC. What else changed markedly enough to explain this?
4) The pattern of increasing health share is quite robust and there is no sign of this slowing down (The apparent assumption that the more middling OECD countries must be at end-stage development or some such is bizarre and largely without merit…. why should this pattern change for even richer countries like the US?)
5) Though new technology is very likely to be the primary means by which countries increase their NHE in the long run, it is probably not a major cause in and of itself. The observed cross country differences are much more consistent with differences in adjusted household consumption and/or disposable income. In a very plausible polynomial specification and +/- year fixed effects, you might notice: the AIC terms barely change; the model fit does not improve markedly; and the year FE are pretty negligible over a 20+ year span. Put differently, if technology itself were itself a substantial cause I would expect NHE conditional on (constant) AIC to shift up markedly with time (i.e., higher intercept) due to constant technological improvements and we largely don’t find this.
6) The best available data pertaining to cross sectional and time series price data suggests actual volume increase explains essentially all of the observed cross sectional differences in NHE and essentially all of the increase in expenditure domestically. (Which is likely in large part use of novel medical devices, drugs, procedures, etc and increasing consumption of those that already exist…. technology, in other words, but if tech is causal why does this volume correlate nearly perfectly with AIC in the long run?)
7) Despite the fact that I’d expect substantial spillovers within national boundaries (redistribution of health spending/resources, standard of care, insurance pooling, etc and thus lower elasticities), there is also substantial variation between states that is quite well correlated with how prosperous those states are.
8) There are also reasonable theoretical reasons to expect to find patterns like these (see this for instance).
> All we really know today is that given the current global economy and current availability of medical technology these other countries are providing medical care of identical effectiveness both at a lower per capita rate and a lower % of AIC.
But these countries are not one big undifferentiated mass and there is plenty of evidence that technology utilization and the like varies strongly with how prosperous a society is in the long term! Despite the fact that the richer OECD countries spend much more than the poorer members (and actually do a lot more in most cases) it appears as if there is no relationship between NHE and life expectancy past a certain point due to rapidly diminishing returns (though I think there are some modest confounds due to higher AIC-> increasingly lopsided caloric balance -> obesity/diabetes/etc and that there are other dimensions to NHE that people care about besides pure mortality measures, like elective surgeries/treatments that may subjectively improve quality of life even though not moving the needle much on mortality).
Norway and Luxembourg, for instance, spend more than twice as much Israel and Malta and see significantly shorter life expectancies. The point being if you’re going to assume away confounds and assume the rest of the world basically has health care figured out you’re going to have to explain patterns like these too. If there spending is actually rational and well managed (for some definition thereof), why aren’t they getting the same bang for their buck? Are you willing to argue that southern europe has better health care systems relative to pretty much all of northern europe (better outcomes and generally much less spending) despite the fact that they are usually less institutionally capable in most dimensions and that it’s just coincidence that their consumption/disposable income is notably less….?
P.S. If you feel I’ve overlooked something I apologize. I gave the matter as much time as I could spare this evening and won’t be following up for awhile, though I’ll read whatever you write back whenever I have down time.
I think the most interesting finding is health care spending is still a standard deviation higher here than elsewhere even after adjustment. Those who argue markets lower costs either have to argue those countries are freer or markets really don’t work that way in health care.
It’s only ~1 SD above trend in the linear specification, but the data are actually substantially more consistent with non-linear growth, i.e., health share of consumption increases increasingly as a function of AIC, in both the broader cross-sectional data (~all countries) and the panel data (OECD and affiliates only). Even when US is excluded from training data, the residual under that alternative (more realistic) specifications suggests the US residual tends towards pretty much smack dab in the middle (and below several countries single payer etc).
https://randomcriticalanalysis.wordpress.com/2016/10/05/predicting-health-care-expenditures-in-oecd-data-with-non-linear-model-specification/
#2: Though I think it very likely the combination of third party payment systems and redistribution of health resources by income (r < ~0 at individual level) stimulates demand rather substantially above a counterfactual where almost everyone pays out of pocket and/or through something like an HSA, these core features are by no means unique to the US and society still decides to maintain and increase funding for these plans we grow wealthier (and/or more willing/able to consume). There does not appear to be much public appetite for a major shift away from third party payment and ultimately the amount we spend is heavily determined by the amount of funding that is allocated towards these third party payers and/or public providers in each period (with a not insignificant amount of lag in policy response to changing economic circumstances).
Within these sorts of systems rationing care is one of the few variables we can realistically move to have a large and sustained impact on expenditures, but it’s not popular and it gets increasingly less popular as societies grow richer. Probably the most successful country from a cost containment point of view was the UK (setting aside other conflicting preferences)… however, after a decade or two of growing public discontent they ‘reformed’ their system to be considerably more generous so that they are very close to the prevailing trend (and still there’s a lot of grumbling!).
#3: I get fairly similar results for adjusted household disposable income which mostly only differs from AIC insofar as it adds actual household savings (+/- gross vs net adjustment for fixed capital consumption). US households quite wealthy overall too in terms of financial net worth. The difference largely comes down to how much GDP is actually available to households as opposed to corporations or government. Differences in fiscal policy and household propensity to consume might contribute to AIC and thus NHE, but it’s far from the only relevant factor.
(Current) GDP does not tell you everything you should want to know about a country. Norway, for instance, is heavily reliant on exploiting finite oil reserves to generate their GDP and once these reserves deplete (or if oil price plummets) their GDP will plummet…. so they engage in considerable consumption smoothing (I submit this isn’t mostly something unique to Norwegian culture/governance but a fundamental difference in economic circumstance). Luxembourg, on the other hand, is a small city-state with a large non-resident work force and huge finance/banking sector, i.e., much of this GDP isn’t really available to domestic households to consume or save….
#4: I am not 100% sure where you’re going with your consumption shares thing, but it’s worth pointing out that there is a great deal of commonality between countries in how they allocate their consumption expenditures at similar levels of AIC and the US is very much on trend here. In other words, it seems likely something more fundamental going on with respect to marginal utility varying between expenditures categories as a function of what countries think they are willing and able to consume (invariant whether some of said consumption is funded publicly or privately) and the changing value of human life. See (longer) post on this tangent here. It may well interact with market distortions and the like, but these features appear to be quite similar. This paper tries to model health consumption more formally and I think it is at least barking in the right direction…