- On being black in Baltimore Olga Khazan, the Atlantic
- What Europeans talk about when they talk about Brexit London Review of Books
- Time to worry James Grant, Weekly Standard
- The English question Paul Harris, Aeon
I’ve been playing around with some data for a paper I have been trying to write about the economic history of Canada in the 20th century. In the process, I assembled the data from the Base de données sur la longévité canadienne regarding life expectancy at birth. Then, I thought that it would be interesting to see how large were the differences between the provinces and how fast did they close. They closed pretty dramatically during the 20th century – see for yourself.
Today is inauguration day. Donald Trump will officially be the 45th President of the United States of America. Many have pointed out that Trump is the oldest president (slightly above 70 years of age). I disagree.
Old is not a “purely” absolute concept. Advances in living standards mean advances in our ability to live longer lives. Not only do we live longer lives than in the past, but at any point in our life, our health is better. Someone who reached 65 years of age in 1900 probably did not have the same health prospects as someone who reaches that age today. Basically, the “quality” of old age has increased over time (see this great book on the economic history of aging). So, when people say “old”, I ask “old as compared to what”.
To meet that test, I took the CDC data on life expectancy as well as soon historical database from 1900 to today. I combined it with David Hacker’s work on life tables in the US from 1790 to 1900 which can be found in this article of Historical Methods. Hacker’s data concerns only the white population. I took only the age expectancy at birth of males. Then, I plotted the age of the president at the time of inauguration as a share of the life expectancy at birth (E0). This is the result:
As one can see, the age of presidents as a share of life expectancy is falling steadily since the early 1900s. In this light, Donald Trump is not the oldest president. In fact, the oldest president is …. drumroll…James Buchanan (1.85 times the life expectancy of white males at birth). Moreover, in this light, the youngest president at inauguration is not Teddy Roosevelt (Kennedy was the youngest elected). Rather, the youngest is Barack Obama followed very closely by Bill Clinton and John F. Kennedy.
I find this post to be interesting as it shows something more important in my eyes: how the poorest in society have done. Presidents have generally stemmed from the top of the income distribution. Over time, the ages of presidents at inauguration (in absolute terms) has not followed any clear trend. The drop seen in the graph above is entirely driven by increases in the life expectancy of the “average” American. In a certain way, it shows that the distance between that “Joe the Plumber” and the “Greatest Man in America” (huh…Lord Acton anyone?) seems to be diminishing over time.
As part of my series of blog post reconsidering health outcomes in Cuba, I argued that other countries were able to generate substantial improvements in life expectancy even if Cuba is at the top. Then I pointed out that non-health related measures made Cubans so poor as to create a paradoxical outcome of depressing mortality (Cubans don’t have cars, they don’t get in car accidents, life expectancy is higher which is not an indicator of health care performance). Today, I move to the hardest topic to obtain information on: refugees.
I have spent the last few weeks trying to understand how the Cuban refugees are counted in the life tables. After scouring the website of the World Health Organization and the archives of Statistics Canada during my winter break, I could not find the answer. And it matters. A lot.
To be clear, a life table shows the probability that an individual of age X will die by age X+1 (known as Qx). With a life table, you will obtain age-specific death rates(known as Mx), life expectancy at different points and life expectancy at birth (Lx)(Where x is age). Basically, this is the most important tool a demographer can possess. Without something like that, its hard to say anything meaningful in terms of demographic comparison (although not impossible).The most common method of building such a table is known as a “static” method where we either compare the population structure by age at a single point in time or where we evaluate the age of deaths (which we can compare with the number of persons of each group alive – Ax). The problem with such methods is that static life tables need to be frequently updated because we are assuming stable age structure.
When there is important migration, Qx becomes is not “mortality” but merely the chance of exiting the population either by death of migration. When there are important waves of migration (in or out), one must account for age of the entering/departing population to arrive at a proper estimates of “exits” from the population at each age point that separate exits by deaths or exits (entries) by migration.
As a result, migration – especially if large – creates two problems in life tables. It changes the age structure of the population and so, the table must be frequently updated in order to get Ax right. It also changes the structure of mortality (exits). (However, this is only a problem if the age structure of migrants is different from the age structure of the overall population).
Since 2005, the annual number of migrants from Cuba to the United States has fluctuated between 10,000 and 60,000. This means that, on an annual basis, 0.1% to 0.5% of Cuba’s population is leaving the country. This is not a negligible flow (in the past, the flow was much larger – sometimes reaching north of 1% of the population). Thus, the issue would matter to the estimation of life tables. The problem is we do not know how Cuba has accounted for migration on both mortality and the reference populations! More importantly, we do not know how those who die during migration are measured.
Eventually, Ax will be adjusted through census-based updates (so there will only be a drift between censuses). However, if the Cuban government counts all the migrants as alive as they arrive in a foreign country as if none died along the way, it is underestimating the number of deaths. Basically, when the deaths of refugees and emigrants are not adequately factored into survival schedules, mortality schedules are be biased downward (especially between censuses as a result of poor denominator) and life expectancy would be accordingly biased upward.
Now, I am willing to reconsider my opinion on this particular point if someone indicates some study that has escaped my gaze (my Spanish is very, to put it euphemistically, poor). However, when I am able to find such information for other Latin American countries like Chile or Costa Rica and not for Cuba, I am skeptical of the value of the health statistics that people cite.
The other parts of How Well Has Cuba Managed To Improve Health Outcomes?
Since the passing of Fidel Castro, I have devoted myself to researching a proper assessment of his regime’s achievements in matters of health care. The more I dig, the more I am convinced that his regime has basically been incredibly brilliant at presenting a favorable portrait. The tweaking of the statistics is not blatant or gigantic, it is sufficiently small to avoid alerting demographers (unlike when Davis and Feshbach, Eberstadt and Miller and Velkoff found considerable evidence of data tweaking in the USSR which raised a massive debate). Indeed, a re-computation of life expectancy based on life tables (which I will present in the new few weeks) to adjust for the false reclassification of early neonatal deaths as late fetal deaths (raising the low infant mortality rate by somewhere 28% and 96%) suggests that somewhere between 0.1 and 0.3 years must be knocked off the life expectancy figures. Given that the variations between different measurements available (WHO, World Bank, MINISAP, CIA, FAO) are roughly of that magnitude, it falls within a very reasonable range of errors. This statistical tweaking is combined with an over-dramatization of how terrible the situation was in 1959 (the life expectancy figures vary from 63.9 years to 65.4 years at the beggining of the Castrist regime). But that tweaking is not sufficient to invalidate the massive downward trend. As a result, the majority of public health scholars seem confident in the overall level and trend (and I tend to concur with that statement even if I think things are worse than presented and the slope of the downward trend is too steep).
Those little tweaks have been combined with the use of massive coercive measures on the local population (beautifully described by Katherine Hirschfeld in what should be an example of ethnographic work that economists and policy-makers should rely on because it goes behind the data – see her book Health, Politics, and Revolution in Cuba: 1898-2005) that go from using doctors as tools for political monitoring to the use of abortion against a mother’s will if it may hinder a physician’s chance of reaching the centrally-decided target without forgetting forced isolations for some infectious patients. Such methods are efficient at fighting some types of diseases, but they are associated with institutions that are unable to provide much economic growth which may act as a palliative counter-effects to how choices may make us less healthy (me having the freedom to eat too much salt means I can die earlier, but the type of institutions that let me eat that much salt also avoid infringing on my property rights thus allowing me to improve living standards which is the palliative counter-effect). With such a trade-off, the issue becomes one of the ability of poor countries to improve in the absence of extreme violence as that applied by the Castrist regime.
Over the next few weeks, I will publish many re-computations of health statistics to sustain this argument as I write my article. The first one I am doing is the evolution of life expectancy from 1960 to 2014. What I did is that I created comparatives for Cuba based on how much living standards (income per capita). Cuban living less than doubled over that 49 years period (82% increase) from 1959 to 2008 (the latest available data from the high-quality Maddison data). Latin American and Carribean countries that saw their living standards less than double (or even decline) are Argentina (+90%), Bolivia (+87%), El Salvador (+68%), Haïti (-33%), Honduras (+71%), Jamaica (+51%), Nicaragua (-17%) and Venezuela (+7%). This forms the low income group. The remaining countries available are separated in two groups: those whose income increased between 100% and 200% (the mid-income group composed of Brazil, Colombia, Mexico, Peru, Uruguay, Ecuador, Guatemala, Panama and Paraguay) and those whose incomes increased more than 300% (the high-income group composed Chile, Costa Rica, Dominican Republic, Puerto Rico and Trinidad & Tobago). I also compared Cuba with a group of countries that had incomes per capita within 20% of the income per capita of Cuba. So, how did Cuba’s life expectancy increase?
Well, using only the official statistics (which I do not fully trust although they are from the World Bank Development Indicators Database), Cuba life expectancy (which was already pretty high by Latin American standards in 1959) increased 24%. However, all other countries – which were well below Cuba – saw faster increases. The countries that had the least growth in Latin America saw life expectancy increase 38% and the countries that were equally poor as Cuba saw life expectancy increase an impressive 42%. Chile, whose life expectancy was only 57.5 years against Cuba’s 63.9 in 1960, also increased more rapidly (also 42%) and it has now surpassed Cuba (81.5 years against 79.4 years) and what is more impressive is that this rate has increased in a monotonic fashion regardless of changes in political regimes (democracy, socialism, Pinochet, liberal democracy) while Cuba’s rate seems to accelerate and decelerate frequently. Now, this is assuming that the figures for 1960 are correct. I have surveyed the literature and it is hard to find a way to say which of the estimates is the best, but that of the World Bank for 1960 is the lowest. There are other rates, contained in McGuire and Frankel’s work – the highest stands at 65.4 years for 1960. That means that the range of increase of life expectancy in Cuba is between 21.4% and 24.2%. Its not earth-shattering, but it makes Cuba’s achievements less impressive (although it is impressive to keep increase life expectancy from an already-high level). But as you can see, more important improvements could have been generated without recourse to such violent means. In fact, as a post that I will publish this week shows, the decline in car ownership from 1959 to 1988 probably played moderately in favor of the increase in life expectancy while the massive increase in car ownership in all other countries played (all else being equal) in favor of slowing down the increases in life expectancy (but being too poor or making it illegal to import a foreign car is not health care and I deem it improper to consider that this accident from misfortune should be praised).
In a way, what I am saying is that the benefit is not as impressive as claimed. Given the costs that Cubans have to assume for such a policy, anything that makes the benefits look more modest should make more inclined to cast a damning judgment on Castro’s regime.
Coming up (I will add the links as they are published) :
- Life Expectancy Changes, 1960 to 2014
- Car ownership trends playing in favor of Cuba, but not a praiseworthy outcome
- Of Refugeees and Life Expectancy
- Changes in infant mortality
- Life expectancy at age 60-64
- Effect of recomputations of life expectancy
- Changes in net nutrition
- The evolution of stature
- Qualitative evidence on water access, sanitation, electricity and underground healthcare
- Human development as positive liberty (or why HDI is not a basic needs measure)
My two blog posts on the health statistics of Cuba have convinced me to try to assemble a research article on the topic of assessing health outcomes under Castro’s regime. My first blog post was that there is a trade-off (the core of the article) that Castro decided to make. He would use extreme coercive measures to reduce some forms of mortality in order to shore up support abroad. The cost of such institutions is limited economic growth and increased mortality from other causes (dying from waterborne diseases or poverty diseases rather than dying from measles).
When I thought of that, I was inspired by Werner Troesken’s Pox of Liberty on the American constitution and the disease environment of the country. I was mostly concerned by direct medical interventions. However, the extent of coercive measures used by Castro go well beyond simple medical care (or medical imposition). Price controls, rationing and import restrictions on many goods could also help improve life expectancy. Indeed, rationing salt at 10g (hypothetical number) per person per day is a good way to prevent dietary diseases that emerge as a complication from overconsumption of salt. That will, by definition, raise life expectancy.
And so will bans on importing cars.
There is an extensive literature on the role that car fatalities has on life expectancy. This paper in Demography (one of the top demographic journals) finds that male life expectancy in Brazil is lowered by 0.8 years by traffic deaths. And traffic has very little to do with the quality of health care services. Basically, the more you drive, the more chances you have of dying (duh!). But, people don’t care much because the benefits of driving outweigh the personal risks.
In Cuba, people don’t get to make that choice. As a result, the very few drivers on Cuban roads have few accidents. According to WHO data, the car fatality rate is 8.15 per 100,000. There is also only 55 cars per 1,000 persons in Cuba. The next closest country is Nicaragua at 93 cars per 1,000 and the top country is Uruguay at 584 cars per 1,000. When you compute reported (rather than WHO estimated) car fatalities per 1,000 cars (rather than persons), Cuba becomes the unsafest place to drive in Latin America (1.46 fatalities per 1,000 cars) after El Salvador (2.22 fatalities per 1000 cars but only 129 cars per 1000), Ecuador (1.78 fatalities per 1000 cars but only 109 cars per 1000) and Bolivia (1.53 fatalities per 1000 cars and only 113 cars per 1000).
The graph below shows the relation between car fatalities per 100,000 inhabitants and life expectancy. Cuba is singled out as a black square. Low rate of car fatalities, higher life expectancy. Obviously, this is not a regression and so I am not trying to infer too much. However, it seems fair to say that Cuba’s life expectancy can easily be explained by the fact that Cubans face stiff prohibitions on the ability to drive. Those prohibitions give them a few extra years of life for sure, but would you really call that a ringing endorsement of the health outcomes under Castro’s regime? I don’t…
On Monday, my piece on the use violence for public health purposes in Cuba (reducing infectious diseases through coercion at the expense of economic growth which in turn increases deaths from preventable diseases related to living standards) assumed that the statistics were correct.
They are not! How much so? A lot!
As I mentioned on Monday, Cuban doctors face penalties for not meeting their “infant mortality” targets. As a result, they use extreme measures ranging from abortion against the mother’s will to sterilization and isolation. They also have an incentive to lie…(pretty obvious right?)
How can they lie? By re-categorizing early neonatal (from birth to 7th day) or neonatal deaths (up to 28th day) as late fetal deaths. Early neonatal deaths and late fetal deaths are basically grouped together at “perinatal” deaths since they share the same factors. Normally, health statistics suggest that late fetal deaths and early neonatal deaths should be heavily correlated (the graph below makes everything clearer). However late fetal deaths do not enter inside the infant mortality rates while the early neonatal deaths do enter that often-cited rate (see graph below).
Normally, the ratio of late fetal deaths to early neonatal deaths should be more or less constant across space. In the PERISTAT data (the one that best divides those deaths), most countries have a ratio of late fetal to early neonatal deaths ranging from 1.04 to 3.03. Cuba has a ratio of more than 6. This is pretty much a clear of data manipulation.
In a recent article published in Cuban Studies, Roberto Gonzales makes adjustments to create a range where the ratio would be in line with that of other countries. If it were, the infant mortality of Cuba would be between 7.45 and 11.16 per 1,000 births rather than the 5.79 per 1,000 reported by the regime – as much as 92% higher. As a result, Cuba moves from having an average infant mortality rate in the PERISTAT data to having the worst average infant mortality in that dataset – above that of most European and North American countries.
So not only is my comment from Monday very much valid, the “upside” (for a lack of a better term) I mentioned is largely overblown because doctors and politicians have an incentive to fake the numbers.