Outcomes are determined by more than health system inputs alone (indeed, it has little to do with health care in the developed world), and costs have much to do with the intensity of health care. Rich countries spend more because they can, not because they need to.
> We spend 17.4% of GDP on healthcare. The OECD average is 9.5%
Health spending is almost entirley explained by income levels, especially in the long-run. The US spends much more because the US is much richer than most and because health spending is highly elastic at a national level.
> and we don't have better outcomes to show for it
Norway and Luxembourg also spend 2x Spain and Italy and don't have more to show for it either despite the fact that they're also much richer, have larger welfare states, etc.
Countries increase health spending because they can, not necessarily because they need to. Evidence strongly suggests returns to health spending are falling everywhere and the US isn't particularly unique in this regard.
Your findings are not mutually exclusive with other theories about the underlying causes of high U.S. healthcare expenditure, such as a chronic lack of price transparency. The more money is in the pot, the greater the incentives to pilfer it, and, in lieu of adequate controls, the more it will be pilfered. The way I see it, your linear regression is between income and aggregate pilferage across all layers of the healthcare establishment (Kafka baklava :) ). I surmise that the accelerating nature of health expenditures (1.8% increase for every 1% increase in income) is due to the rapid inflation of disposable income relative to overall income at the higher levels. Disposable income and the saved wealth accrued from higher disposable income over time are more readily pilfered. I should say “otherwise disposable,” because the healthcare industry takes a progressively larger chunk of that and makes it de facto indisposable. I say this with no irony: a linear regression between income and amounts extorted during kidnapping, controlling for other variables, would show similar results.
> Your findings are not mutually exclusive with other theories about the underlying causes of high U.S. healthcare expenditure, such as a chronic lack of price transparency.
My findings strongly agitate against the notion that high US health spending is a product of idiosyncratic features of our system. Presumably most of these critics believe we'd spend much less if only our system looked more like other countries, but my evidence indicates we'd spend very similar amounts in the long run regardless. Further, we'd likely have similar outcomes and many other similar healthcare attributes (prices, intensity, health worker density, etc). Most of the things about US healthcare people believe to be important and unique (i.e., not explained by income) just aren't.
> The more money is in the pot, the greater the incentives to pilfer it, and, in lieu of adequate controls, the more it will be pilfered
I wouldn't argue there's no "pilfering" or that more money doesn't create more opportunity for this, but the high income elasticity likely has little to do with pilfering. Where are these ill-gotten gains going? I certainly don't think you'd have much success in showing this if you look at, say, the growth in physician incomes, pharma/biotech industry profits, and so on. The data are much more consistent with mundane explanations like rising technological sophistication, higher intensity, and so on (ultimately much of this being driven by some combination of patient/family demand and providers' "spare no expense" approach to caring). I mean, if you look at the economic data it's quite obvious most of the increase can be arithmetically attributed to a swelling of health workers (density or share of workforce) and that most of these workers have lower-to-middle income levels (especially on the margin).
> I surmise that the accelerating nature of health expenditures... is due to the rapid inflation of disposable income relative to overall income at the higher levels.
I'm not sure what you mean by this exactly, but a better, more parsimonous way to understand high income elasticity is that higher income countries are usually inherently more productive countries. We can spend substantially smaller shares of our income on food, clothing, shelter, and other "necessities" because we are able to produce these things so much more efficiently (or otherwise procure on the market) than we did decades earlier or than OECD countries of more significantly more modest income levels while still consuming more of these things in real terms. This frees up resources to be spent on higher order wants like health, education, recreation, culture, and so on. Many of these growth areas, meanwhile, are inherently subject to less productivity growth, meaning prices tend not to fall relative to incomes nearly as quickly as we observe in other sectors.
> I say this with no irony: a linear regression between income and amounts extorted during kidnapping, controlling for other variables, would show similar results.
I doubt that's true, though we're talking about the total spending (kidnapping ransom) per capita here. It's pretty clear the price per transaction (as in, health inflation) explains very little in US time series or cross-sectionally. Now maybe if kidnapping started to become a high amenity affair in developed countries (presuming this sort of thing happened with measurable frequency here) and 25% of the population worked delivering these services.....
Nothing is perfect, but most experts believe this has little to do with healthcare today because healthcare interventions tend not to be effective causes of long-run weight loss and most countries aren't doing enough of the stuff likely to have large effects (e.g., surgical interventions) to explain much of the variance. Even if you could argue it might explain something, say 0.5 mean BMI points, other factors are clearly highly important. Cultural * and genetic factors are likely to play a significant role amongst high-income countries. Further, obesity rates rise with time and income levels despite higher health spending.
> socialized healthcare does a better job of controlling obesity with preventative health measures
evidence?
> As it is now in the U.S., patients only go to medical professionals when there is a problem
The US spends more on preventive medicine than almost any other country, though preventative medicine generally has very-small-to-modest effects on outcomes and rarely, if ever, saves money (usually quite the other way around)
~ RCA
note: * some of these "cultural" factors may be residual economic influences... the US escaped the malthusian trap long before almost all other high-income countries and this may have latent effects on attitudes towards food, diet, etc)
> The obesity explanation doesn't pass the sniff test. Canada is 12% less obese than the U.S., but it spends 6% less of its GDP on healthcare.
It's my blog (RCA). My argument is that obesity substantially explains US health outcomes in relation to other countries. I never claimed obesity is the cause of high national health spending (as in, "inputs"). To the contrary, I have consistently argued US health spending is well explained by its wealth (technically income levels).
To a first approximation, national health spending is entirely explained by the average house income level in the long run. While time, healthcare technology, and other factors are assocatied with rising spending, these changes are ultimately very well explained by changing income levels.
Amongst high-income countries, a 1% increase in income is robustly associated with a long run increase of about 1.8% (it's highly elastic).
The US spends more than Canada because it's still a much richer country (which isn't to say Canada isn't a nice place!).
> That is beyond the realm of believability, even if I introduce the other population-induced causal factors which you implied but didn't specify.
Again, I never said this, but other population health risk factors such as age structure, disease rates, and the like are of negligible significance when it comes to long run aggregate spending. Such factors may be highly predictive within countries and may have some say in the short run (within budgetary constraints), but in the long run national picture the evidence suggests these factors amount to little more than noise. National household income levels trumps everything.
> Additionally, the government would be more invested in the population's health under a single-payer model.
US government programs, namely Medicare and Medicaid, spend more on healthcare than most other high-income countries do in total (even more so comparing public-to-public). Just how much more incentive do we need before these magical effects kick in? Higher health spending predicts higher obesity rates in time series and cross-sectionally (though this is likely ultimately mediated by long-run income levels and by time).
Where is the evidence that these programs have large, sustained effects and are cost effective? Most data indicate these programs have negligible effects in the long run and they almost always cost more than they save (which isn't to say we shouldn't necessarily do it, but the economic rationale is v. weak).
> a) many (perhaps the majority, don't have time to source now) of the bankruptcies in the US are due to medical issues and spending
The vast majority of medical bankruptcies have nothing to do with the cost of medical care, but the disruption to career/income flow imposed by illness. This is clearly a problem in other countries as well.
> The rate of growth of US spending on healthcare....far outstrips the rate of growth of money
This is true in other OECD countries too. Further, this doesn't mean what you think it does. As our productivity rises, the share spent on consumption categories with high productivity growth (increasingly low relative prices) can decline, which frees up spending to be spent on health and other areas subject to less productivity growth (the majority of the expenditure growth corresponds to rising real health consumption tho)
> The same medicines by the same manufacturers and often the same production lines ... often cost 10-100 times more in the US than they do in other places
One might be able to find outliers of this sort, but that clearly doesn't reflect anything close to central tendencies (mean, median, mode, etc), especially when compared (accurately) to other high-income countries. Richer countries, like the US, generally pay relatively higher prices.
The US may pay a somewhat higher premium, but there are tradeoffs here vis-a-vis incentivizing innovation in the long run. It's also not widely appreciated that the US pays markedly less for generics....
I'm not a language zealot (most languages/frameworks come with pros and cons) and I use R and the tidyverse quite regularly, but "performant" is not a word I'd associate with the tidyverse. I'd be surprised if it wasn't slower than most alternatives (can't claim to have systematically benchmarked it, still....), even if it's often easier to use and usually "good enough."
I (RCA) have been meaning to dedicate a blog post to this topic exclusively and quantify the root causes in considerable detail.
In short, yes, I believe globalization and, specifically, multinational corporations (affiliates of foreign owned corporations) are the primary cause. However, in regression terms, this has less to do with net income flows into the US than net income flows out of several small high GDP/person countries. The US is a very large, very rich country whereas many of these countries are much smaller so it doesn't take much to have outsize affects on them. Even if 100% of this ultimately accrues to the benefit of US households, (which I don't think is quite accurate) the effect in % terms is vastly different.
Although some of this can be directly observed as primary income flows in the current year national accounts (dividends, rent, etc), some of it shows up as gross savings or disposable income in the foreign affliates, i.e., it's equivalent to retained earnings. Over the past decade or two non-financial corporate savings have increased massively and these savings are largely uncorrelated with domestic (capital) investment, i.e., it's almost entirely financial and it's largely leaving these countries in the form of net lending. These things also influence the calculation of (GDP) PPPs and cause other headaches.
Of course, there are also other reasons why GDP misleads. For example, Luxembourg has a very large non-resident workforce. Something like 50% of their workers live in neighboring countries (varies year to year), meaning ~50% of aggregate employee compensation goes home (cross border) to Germany and the like. Then there are petro-states like Norway whose income flows are inherently temporary (finite amount of natural resources to extract) and highly volatile, meaning they can't consume out of their measured GDP like most other countries. They need to practice massive consumption smoothing if they don't want their standard of living to crash in the not too distant future.
Long story short, GDP was never intended to be an indicator of material wellbeing and the household perspective (consumption, disposable income) are better measured and more reliable. One might try to throw a bunch of variables in to counteract the many issues imposed by GDP as a proxy for the household perspective, but why bother?
~ RCA (sorry for typos, grammatical errors, etc... limited time to comment and would rather focus more effort on blog)
1, I can fit the US on a linear trend amongst high income countries.
2, there’s no necessary reason why increasing health share with rising real income is unsustainable. We can and have increased share spent on health while increasing real expenditures across the board.
3. I touch on some reasons why this may curve up and then eventually flatten out.
4. No, not everything fits on log-log slopes and us is very close to the trend.
2. Asymptotic growth towards 100% certainly sounds unsustainable.
3. You mention Baumol’s cost disease and the proportion of income spent on services, which is not so bad.
4. Log-log overfitting is a well-known phenomenon and a low deviation (particularly at the edge of the graph) doesn’t make it go away.
5. This claim, tucked between historical spending and nurse salaries, underpins much of the thesis, but is curiously unsubstantiated:
>Nor do we tend to find results consistent with this in wages, profits, and other proxies for (or presumed causes of) such issues in these sectors. On the contrary, the reliable statistics for healthcare (at least) shows prices have fallen relative to average nominal income and that most of the increase is therefore explained by rising real consumption (quantities per capita).
Healthcare prices are kind of the whole point here. The lower third of this country can’t afford essential care. Not only that, but we’ve seen the prices, and they’re ridiculous.
So if you have data about healthcare prices not being out of order, that seems a lot more relevant— and less cherry-picked — than the history of household spending on food consumption.
> 2. Asymptotic growth towards 100% certainly sounds unsustainable.
"Sounds" isn't an argument and I'm explicitly arguing the slope is likely to flatten, eventually. The point the income elasticity of health expenditure can be (is) well north of one and we can (and do) consume more of everything else at the same time.
That isn't a thing. You can argue this specification minimizes the residuals at the high end if you want, but it's very likely to the correct modeling decision, it's bog standard in economics, and the US residual ~= 0 (certainly not notably high). It's also pretty obvious health expenditures are increasing in % terms and that failing to log-transform results in particularly poor model performance with constant slope out of sample.
> The lower third of this country can’t afford essential care.
The lower third of the country consumes approximately the same amount of care as the rich, as in other high-income countries, and the socioeconomic gaps in other countries are likely comparable to even larger (depending on how measured).
To the extent there are real and ultimately consequential issues with affordability for some small segment of our population, these aren't likely to be explained by aggregate costs or prices so much as by narrow details that we can tweak, i.e., without requiring massive change, should the political desire exist to do so.
> seems a lot more relevant— and less cherry-picked — than the history of household spending on food consumption.
Pardon me, but I was engaging with someone that was arguing this expenditure growth implied starvation and you're making very similar (wrong) arguments. Whether you appreciate it or not, the role of general increases in productivity, the source of real income growth in the long run, and differences in the rate of productivity growth in different sectors, which we are clearly reflected in prices, is very much on point. Food production is simply a way to make this concrete for people that struggle with abstractions like price indexes and relative prices.
So I think it's interesting here that you make no mention of the explanation for HCE increases provided in the "consensus view" link.
>In health care research, the impact of medical technology on health care cost increases has always been a great unknown. Yet 81 percent of the leading health economists agreed with the statement, “The primary reason for the increase in the health sector’s share of GDP over the past 30 years is technological change in medicine”.1Growing attention to the role of technological change in driving growth in health spending, and to the costs and benefits associated with new medical innovation reflects an acknowledgement of the long-term dilemma posed by historically unsustainable rates of growth in medical costs, combined with an increasing consensus that technological advance is a major factor in driving this growth. The current acceleration in health spending growth - following the quiescent period accompanying the spread of managed care - brings troubling implications for the long-term viability of our current system of financing and provision of health services. Understanding the magnitude of technology’s historical contribution to growth in costs is vital to the analysis of the future path of medical spending. Of course, in most areas of the economy a rapid pace of technological advance is regarded as a good thing. That this is not the case for medical care reflects a second point of consensus. Throughout much of history, imperfections in medical care markets have failed to provide incentives for the cost-effective provision of medical services, encouraging the development and diffusion of innovations beyond the point that would prevail under competitive market conditions. Low out-of-pocket costs for medical care due to insurance coverage, combined with patients’ lack of full information on the services they consume encourage the provision of medical care to a point where the marginal benefit of treatment to the patient is small relative to its marginal cost.
You can hand-wave this as "increased consumption", if you want, but it is:
- recognized as an anomaly
- considered a point of concern
- likely to lead to cost reductions if fixed
In other words, this is precisely the sort of phenomenon you are arguing does not exist in US healthcare!
> Yet 81 percent of the leading health economists agreed with the statement, “The primary reason for the increase in the health sector’s share of GDP over the past 30 years is technological change in medicine"
I've highlighted the role of technological change on my blog before, but technological change is a major proximate cause. The root cause is income growth. Countries are chasing these technological advancements in direct proportion to their income and the degree to which they can afford them (as prices fall, countries with lower real incomes are more able to afford technologies the US and other rich countries had long before, but the frontier has long since moved on....). If you throw year fixed-effects or a time trend into analysis the coefficient on time (a proxy for tech chg) is very modest and the income effects are virtually identical.
