The Power and the Money

Now my final word on the demographics of the War of the Triple Alliance, really!

 First, Brent is retired and the relatives do not live in affected areas. Thank you for the concern!

Second, Paraguay. While I await the coming storm of comments and questions (hah!) here is the population pyramid for the 1870 data with the estimated undercounts from 1899 added in:

Devastation among males (55%!), but little sign of a generalized crisis. Which is actually somewhat surprised: I would have expected blowing that big a whole in the labor force would have led to severe general famine.

Note that these adjusted figures are a clear undercount. First, the 1899 figures include only those who survived to 1899; some people who were not counted in 1870 passed away in the interim. Since males have shorter lifespans than females, this will also increase our projections of missing males. The 47% figure from the 1886 census is a more reasonable upper bound.

Second, the total number is too low. It comes to about 176,000. That implies an annual growth rate of 3.8% between then and 1886, which is impossible unless there were lots of Paraguayan refugees in Argentina and Brazil in 1870. There almost certainly were ... but that means that the 1870 census cannot be used to impute total deaths. 

Here is a table of plausible populations given an 1886 population of 325,000:

1870 pop at	3.0%	201,067
1870 pop at	2.5%	217,813
1870 pop at	2.0%	235,954
1870 pop at	1.6%	251,549

The irony? The War of the Triple Alliance was still devastating. It was on the same scale as the Greek War of Independence, the French conquest of Algeria, World War I in Serbia, and the Great Patriotic War. It was more destructive than the Carlist Wars, the Ten Years War, the Russian Civil War, and the two Congo Wars. It was more than twice as bad as the Napoleonic Wars were for France, inflicted over half as much time. It was a horrible event.

What it was not was singular or unprecedented. Which says terrible things about human beings.

Paraguayan demographics in the late 19th century

 Here is my final word on the demographic impact of the War of the Triple Alliance. Let’s start with the 1899 population pyramid. There is some guess work here, because Asunción accounted for 10% of country’s enumerated population of 480,000 but provided no age breakdown. These figures heroically assume that the capital had the same distribution as the rest of the country, which is almost certainly wrong. (Cities skew older in general; more so in the age before modern urban sanitation.) But the numbers are what we have.

What do they show? Unfortunately, the age groupings make it hard to see whether there are any signs of a general demographic disaster (that is, one affecting both sexes) during or right after the war, but there is nothing particularly obvious. With two exceptions, each cohort looks more or less the size you would expect in a growing mostly-rural population. 

The exceptions are the 1854-63 and the 1882-84 birth cohorts. The number of females in the first cohort is slightly smaller than you would expect given the size the cohorts around it. That population was aged 1-10 at the start of the war, so it is possible that we are seeing some effect of disease and malnutrition. The problem is that the cohort after that includes everyone born between 1864 and 1881: war-related mortality among those born in 1864-70 would be swamped by those born in the postwar decade. 

The other exception is the 1882-84 birth cohort, which is much bigger than you would expect for both sexes. I suspect that the reason is a head tax on adults or some sort of labor draft that affected both sexes: children who were actually aged older than 18 in 1899 were reported as aged 15-17 instead. (There may be an alternate explanation, but it is unlikely to involve the war.)

Now let’s take a glance at the 1886 pyramid.

What I’ve done here is add back the missing birth cohorts from the 1899 census. The categories did not easily overlap, so I have divided up the cohorts from the 1899 census equally among all the years inside that cohort. (This introduces biases, although I think they run against the hypothesis that the war had no effect.) 

Why were the undercounts so large? Well, one possibility is evasion. Another is a sloppy census. But commentator Thomas Masterson hits on a third hypothesis: temporary migration to Argentina. Argentina was booming during the 1880s and getting their from Paraguay was easy. Paraguayan migrants, however, had to compete with the waves of Spaniards and Italians pouring into that country. The Paraguayans would face racial and language discrimination (Guaraní the primary language of most locals at the time), which would not be conducive to permanent settlement. Finally, the 1899 census identified Paraguayans by nationality, not place of birth, and so Paraguayans born in Argentina but who moved back would be counted in the 1899 census as local. 

What about the impact of the war? There are clear signs of mass casualties among males: the 1886 male population is a full 41,000 less than might have otherwise been expected ... although that calculation does not account for the fact that male mortality is generally higher than female. Excess male mortality comes to 47% of the population of the relevant age group, which is consistent with the idea that about one out of every two Paraguayan males perished as a result of the war. (It also implies that almost the entire male population served, but that is consistent with Paraguayan records and the overwhelming reports of children and elderly being pressed into service.) 

That said, it is a little perplexing that there is also excess male mortality among children born during the war ,i.e., too young to have been child soldiers. There are reasons to think that male children might have been more susceptible to famine and disease, but that smacks a little of post hoc reasoning to me.

Looking at the female half of the distribution, there are some signs of higher-than-expected mortality among the cohorts born during and immediately after the war, but they are not pronounced. Paraguay experienced disease and famine (the latter partially as a result of a huge diminution of the agricultural labor force for military service) but not outside the realm of the historical experience of 19th-century wars.

Finally, it should be noted that even with my adjustments the 1886 number is still an undercount. To be fully consistent with the 1899 number the population would need to grow by 3.0% per year. That is really outside historical experience in the latter 19th century: Costa Rica and the Dominican Republic were the highest (without mass immigration) at around 2.6%. Most countries were around 1.5% or less.

Criticisms, suggestions, new information and additional questions, please! (I know that Randy is interested in why Paraguay survived as an independent nation ...)

How not to do history, Paraguayan edition

 Commentator J.H. asks what I think of the Whigham-Potthast methodology for estimating Paraguayan deaths in the War of the Triple Alliance. Coincidentally, the Whigham-Potthast article was exactly what I planned to be the subject of this post!

But first, let me say that I agree with J.H.’s point that had the country really lost 90% of its male population, then we would have seen bigger effects on its culture and literature than we do. Let me also say that careful researchers (like Reber) have tried to adjust for Paraguay’s territorial losses, but regardless those losses affected very little populated territory.

Which brings us to the work of Thomas Whigham (U-Georgia) and Barbara Potthast (Bielefeld). They uncovered an 1870 population survey taken at the tail end of the war. The survey was the work of the Provisional Government established by the Allied countries, who were not yet prepared to simply partition Paraguay between them. It isn’t entirely clear why the Provisional Government wanted the survey, but the answer is likely taxation: the government asked that information be gathered on the acreage sown with various crops in addition to the population data. 

Now, the survey wasn’t really a census: rather, the government asked the political chiefs and justices of the peace in the various counties to report the data. Age groups were divided into children, adults and elderly, with no further elaboration. In addition, the survey omitted the cities of Asunción and Pilar. But it was data, and data are always useful! Consider what we were able to do with the highly flawed 1886 census.

So what did Whigham and Potthast do with their data? First they added estimates for Asunción and Pilar. The Asunción data came from an 1872 British diplomatic report; the Pilar data came from an 1871 newspaper article. Then they added up all the reported numbers of people.

And ... uh ... well, that is what they did. They got a total of 116,351. To that they added a guess of 50,000 people in non-reporting districts. That made a high estimate of 166,351. They then divided that number by their estimate of a prewar population of 389,000 to 457,000. Ergo, a population decline between 57% and 69%.

Do I have to explain why this is unlikely?

Oh, all right. The 1870 numbers do not jibe with other data. First, the 1870 survey reports 39,334 children. Whigham and Potthast take “children” to mean age 12 and below; which would place them as born between 1858 and 1870. In 1886, the number of the children born between 1856 and 1871 was 78,605. That is a difference of more than 39,000 people. A simple adjustment to remove 1856, 1857 and 1871 would cut that difference to 32,000, but it would still mean that more than 60% of the Whigham-Potthast upward adjustment of 50,000 would have to consist of uncounted children for their results to match 1886.

It is harder to match other cohorts. In 1870, adults were defined as anyone born between 1820 and 1857. They totaled 29,310. In 1886, people born between 1816 and 1855 totaled 52,584. Even people born just between 1836 and 1855 totaled 40,648. That is an additional 11,000 people at minimum, which would when combined with the missing children just about eat up their 50,000 fudge factor.

The Whigham-Potthast numbers are still remotely possible but ...

... the second problem is that we know that the 1886 census was also an undercount. How do we know that? Well, the 1886 census reported 28,113 people born between 1882 and 1884. By 1899, that cohort had grown to 43,907 people, all Paraguayan-born. That is, shall we say, not possible unless the 1886 count was too low. The same applies to later cohorts. The 1886 census found 80,476 people born between 1866 and 1881; the 1899 census reported 122,955 people born between 1864 and 1881. Unless 1864 and 1865 were rather astounding fertile years (which would be odd, considering as they were the first two years of the war) then the 1886 count once again has to be too low. (This problem does not apply to older cohorts; it seems that 1886 was particularly bad at recording children and young adults.) 

In other words, 1886 suffered a minimum undercount of 58,000 relative to 1899; and 1870 suffered a minimum undercount of 43,000 relative to 1886. That is a total minimum undercount of 101,000. Lopping off the 50,000 fudge factor leaves us with an additional 51,000 missing people ... for a new minimum 1870 population estimate of 217,000.

To double-check, let’s adjust the 1886 numbers using the 1899 data for the undercounted cohorts. That would give us a minimum population of 290,061. (I feel ridiculous using more than two significant digits to report these numbers, but whatever.) That implies a growth rate of 3.5% per year between 1870 and 1886, which is really not possible in the age before antibiotics. In fact, Paraguay never mustered growth much above 1.7% per year; the figure of 3.1% between 1886 and 1899 is an artifact of the 1886 undercount. 

To triple check, let’s match the 1870 data to the 1873 data. (Which actually reported numbers for 1872.) Now, we do not have original manuscripts for 1873, but I am not clear as to why we should discount its existence given that it was reported in British consular reports and Argentine newspapers. I am certain that the 1873 data are terrible, but that also applies to 1870. Anyway, the 1873 data report 86,079 people born between 1858 and 1872, which about 46,000 more than the 39,334 reported born in 1858-70 in the 1870 survey. (Jan Kleinpenning of the University of Nijmegen got to this point first.) 

Finally, their 1864 baseline is almost certainly too high. It ignores the 1864-67 survey data. (That would be point #4 at the link.) The lowest estimate that they have is barely possible, but still not that likely. 

I hate to make the calculation, but if we combine the high estimate of the 1864 population (370,000, assuming an average household size of seven people) with the low estimate of the 1870 population (217,000) we get a roughly 41% decline over the six years. Not all of that will be excess mortality, of course. Some will be international refugees, others away from their homes in Paraguay (for example, men still in the field who have not returned home), others undercounted, and some normal deaths that were not compensated for due to low fertility during the war. In other words, 41% is a meaningless figure ... but it does at least serve as a bracket. 

A more realistic 1864 estimate uses an average household size of 5.5, which was typical for Chile and Mexico at the time. That would give a prewar population of 292,000. A population size of 292,000 implies 1.8% growth between 1864 and 1899 (and 2.2% growth from 1870 to 1899). Combining that figure with an 1870 population pf 217,000 pushes the total population fall to 22%. That is still, of course, a meaningless figure ... but it is the equivalent calculation to the more dramatic claims.

In short, the Whigham-Potthast method is a good one, but in this case it relies on bum data. The war was clearly catastrophic, but a 57%+ genocidal massacre it was not. 

The existing debate on the Paraguayan War

 After I finished the legwork that went into the last two posts, I discovered two other scholars had gotten there first. In 1988, Vera Blinn Reber took on the idea that Paraguay had lost more than half its population. Her evidence fell into four buckets:

1. We have four decent censuses: 1792, 1846, 1886 and 1899. The intercensal growth rates are 1.7%, 0.8%, and 3.1%. The censuses distinguish between native-born and foreign-born populations, so we can strip out immigration, which was anyway near negligible. The numbers given imply a slowdown between 1846 and 1886, but not a catastrophe;
2. The 1857 census that gave a population of 1.3 million does not exist. It appears to have been a Stalinesque propaganda ploy by the Paraguayan government to frighten its enemies. 
3. Once the war broke out, the Paraguayan government undertook a series of annual household surveys for the purposes of taxation and conscription. If we assume a household size of 6.98 people (which is very high by the standards of contemporary Latin American countries, but about what the 1846 census showed) then the population was 370,073 in 1864. It declined to 304,824 by the end of 1867, but that could have been evasion combined with the fact that much of the male population was at war.
4. Finally, biannual figures on crop production (Paraguay had a remarkably well-organized war economy) show no declines between  1863 and 1867. 

While she effectively debunked the genocidal claims, her conclusion that Paraguay lost most likely 8% of its population is still largely conjecture. For example, roughly 80,000 Paraguayan men cycled through the armed forces during the war. She posited that it was unlikely that casualty rates could have been much higher than 24,000; roughly 30% of everyone who served. 

I am not so sure. For example, the standard figures for Confederate military mortality (including disease) in the U.S. Civil War ranged from 23% to 38%. But when J. David Hacker (Binghampton) used census records to estimate excess male mortality for the 1860s, he found numbers that were more in line with mortality rates of 33% to 54%. Considering the sex imbalances in the 1886 census, I do not find 50% mortality among Paraguayan soldiers to be inherently unbelievable. 

Her doubt comes from the fact that the size of the Allied forces arrayed against Paraguay has been greatly exaggerated. For example, there are claims that Brazil fielded an expeditionary force of 100,000. The problem is that we know that it took the Brazilians a full 49 steamships to move a single brigade consisting of 5,445 enlisted men up the Paraná river and the Brazilians never possessed that sort of logistical capability. That said, however, even if Allied forces never much exceeded 43,500, they could have caused much higher casualties upon the Paraguayan defenders through disease and malnutrition inflicted via the astute use of strategic interdiction. 

Given the evidence from 1872 and 1886, I would lean towards her high estimate of 18% total losses (including a net refugee outmigration around 3% of the prewar population), simply because the military mortality rates that she considered too high for a 19th century war are not in fact outside the bounds of historical reality. Disease made many 19th century wars more dangerous for soldiers than their early 20th century equivalent; in the second half of the 20th century advances in body armor, medevac technology, and trauma medicine produced a second prolonged decline. Consider, as evidence, the below graph of the wounded-to-killed ratios in various wars (from the amazing work of Tanisha Fazal of Notre Dame, available here):

An 18% loss, however large, is still a far cry from over half. 

In 1999, two scholars attempted to resurrect the vision of a devastated Paraguay using evidence from an 1870 population survey. Their attempt was terrible. So terrible that it is almost not worth discussing. But I will! So stay tuned.

Meanwhile, any thoughts before we turn to the political implications of the conflict?

The plot thickens about the War of the Triple Alliance

 Paraguay appears to have attempted a desultory census on January 1, 1873, three years after the end of the war. What we have available is not as comprehensive as from 1886, but there are some age data. The resulting population pyramid both confirms and contradicts the data from 1886:

It confirms the data from 1886 because it also shows a giant skew in the sex ratio for the population born between 1848 and 1857; that is to say, the population aged 7 to 16 at the start of the war and 13 to 22 at its end. So far, so good.

The problem? Well, consider this table:

Birth cohort	Males	Females
1848-57 in 1872	15,083	45,567
1846-55 in 1886	6,420	18,697

That is a whole lot of decline over 14 years! (57% for males, 59% for females, although the cohorts are not exactly comparable.) Is that an indication of the terrible coverage of the 1886 census? If it is, other age cohorts should experience similar declines. 

Birth cohort	Males	Females
1858-72 in 1872	39,507	46,572
1856-76 in 1886	43,202	55,447

and

Birth cohort	Males	Females
Pre-1847 in 1872	13,663	60,678
Pre-1846 in 1886	13,125	42,395

They do not. In fact, given expected mortality, measurement errors and the fact that the cohorts are not equal, the 1848-57 and 1846-55 cohorts stack up remarkably well between the censuses. (The implication is that the 1886 count was a little better for that cohort.) The fit for the pre-1847 and pre-1846 cohorts is worse but still decent; it implies that men aged 25 and above were undercounted in 1872 relative to other groups because it is not possible for their numbers to have declined so little over 14 years. 

The weirdness of the first cohort poses a problem. At face value, these numbers imply that something in 1872-1886 dramatically reduced the measurable population aged 15-38 while leaving other groups untouched. (15 is the youngest member of the cohort in 1872; 38 is the age of the oldest member in 1886.)

I have no idea what that could possibly be. Something that disproportionately kills prime-age adults? Unlikely. Something that makes prime age adults more desirous of avoiding the census in 1886 than they were in 1872? Paraguay was still under foreign occupation in 1872, so its possible that they distrusted their own government more than the occupation authorities. But it seems unlikely.

In short, its a mystery. It seems as though the war ripped a chunk in the fighting age male population of the country. And there is (weaker) evidence that epidemics in the late stages of the war reduced the overall population. But as for everything else the numbers seem far too unreliable to make an estimate.

As you might expect, some scholars have been there before us. A review of their evidence will be forthcoming. 

Did Paraguay really lose 70% of its population in the Great War?

 Apologies for the absence. Work and travel, which may generate some future posting. But right now, some promised history.

The War of the Triple Alliance (1864-70), in which Paraguay went to war with Argentina, Brazil and Uruguay, could fairly be considered one of the world’s most obscure conflicts. But it nonetheless comes up in the most unexpected quarters, usually as an example of the futility of war. The claim to fame is the loss of 90% of Paraguay’s male population (and some unknown but significant portion of its female population as well) making it one of the first genocidal conflicts of the modern age.

So what do we know? We will be looking at this in a series of posts, but let’s start with the Census of 1886. As censuses go, it is completely unreliable in its count. For example, the only way to jibe it with the 1899 census is to assume natural population growth of 8% per year (page 10 here) which is, to put it mildly, not possible.

That said, a population pyramid may start to shed light on the question of wartime population loss. If one assumes that either (1) the undercount was equal across ages and sexes; or (2) the undercount was randomly distributed across ages and sexes, then sudden shifts in the relative size or sex ratio of various cohorts may contain useful information. And so, we present to you the 1886 population pyramid for Paraguay, broken down by birth cohorts:

Something appears to be going on. First, the size of the 1866-1876 cohort is unusually small. This could be a random undercount ... or it could indicate very high mortality among that cohort. The first four years of the cohort coincided with the end of the war and large-scale cholera outbreaks; mortality could have been unusually high.

Still, high child mortality due to war-related disease is not what people commonly think of when they discuss the mortality of the War of the Triple Alliance. Rather, what they think of is massive male mortality from combat. (That would include disease, of course, but disease caused by terrible field conditions rather than a general epidemic.) Is there evidence for that?

Yes. As shown above, sex ratios reported in the 1886 census declined monotonically with age. It is of course likely that heads of household (e.g., males above the age of 30) disproportionately avoided census takers. (Widows were rather less likely to be taxed or drafted.) But that applies to other censuses in other countries as well, and those censuses do not show these sorts of extreme skews in their adult sex ratios.

In short, the 1886 census cannot tell us how badly the war impacted the Paraguayan population, but it can tell us that the violence was significant. More to come.