The answer is deceptively simple: because the reproductive chances of a person “on the spectrum” have changed. In other words, the (small) proportion (or incidence rate) of autistic human beings has grown, and will continue to grow. The primary conjectures I have seen bandied about are:
1. Greater awareness produces more diagnoses (and, usually, “the proportion of autistic humans is constant,” sometimes “because the standards are changing.”)
2. Chemicals (vaccines, pesticides).
What are we trying to explain? Here’s a graph depicting diagnoses in the US.
As cool as that graph is, it has an unhelpful temporal domain. Hays Golden, a PhD student at University of Chicago, has a job market paper (Assortative Mating and Childhood Autism) that not only has a much better temporal domain, but is very cool from just about any angle one would want to take (h/t the Freakonomics blog). More on that in a moment. If we want to understand why autism is rising we really want to think about this over, say, the past two or three centuries, and we are going to want to think about change over different periods of temporal aggregation. The graph above uses the year, and doing so produces a really cool looking graph that gets folks attention. But we do not want to stop at the year: we also want to think about change over half-century chunks, and decade chunks as well.
Why are the temporal domain and units of temporal aggregation a big deal? The reason is that we have a fact we want to explain: the rise in autism rates, perhaps as represented in that graph above (let’s go ahead and assume that the positive growth (not the rate itself) depicted in that graph is representative of many other annual-aggregation graphs we might create since, say, 1940, in whatever country one might wish to examine). Our task, then is to not only come up with an explanation consistent with that face, but to then ask what other facts should we also see if our account is helpful. Then we can see whether our explanation not only accounts for the fact we want to explain (the rising incidence of autism), but all of the other facts that should also be true if our explanation holds. Then we can compare our account to all of the others and see which best accounts for all of the things it should. Though this blog post is not a scientific paper containing rigorous analysis of data, I hope to persuade you that the most plausible account for the rise in autism during the 20th century is due primarily to the increased reproductive success of autistics (which is to say, folks like my dad and myself).
Let’s assume that autism has been randomly produced among a very small portion of human beings for, say, five or more centuries. As we move back in time, more and more human beings were born into societies where face-to-face relationships were paramount: governments that issued ID cards did not exist, and most people knew well those with whom they interacted. I trust you will agree that it is reasonable to believe that the reproductive chances of those few who would today be classified as “on the autism spectrum” were dramatically lower than those who would not today be “on the spectrum.” To the extent that nobody “on the spectrum” was able to reproduce, the proportion of people born “on the spectrum” would remain constant. But, if the chances that someone “on the spectrum” was able to find a mate and reproduce grow, then over time we would see considerable growth in the proportion of autistic human beings. So our question becomes: is there some reason to think that during the 20th century the reproductive chances of autistics changed (dramatically)? It turns out that the answer is: yes.
So, what changed?
1. Incomes for Engineers and Mathematicians
2. Drop in communication and transportation costs
Hans Asperger first used the term autism in 1938 and Leo Kanner followed suit in 1943. To be sure, this is in no small part a story about greater awareness (#2 above): we cannot talk about the rise of something that people have not yet decided to classify and measure. And if we go back 50 or 100 years, well there was little being measured well in the realm of mental difference (indeed, by contemporary standards, it wasn’t being done well mid-2oth century). Nevertheless, it seems reasonable to suspect that by the mid 20th century there was already some increase in the reproductive success of people with autism, at least (and perhaps only) among males in countries in Western Europe (and its colonial Diaspora), the US and Japan. Why? The economic value of engineers had increased dramatically since the industrial revolution: in the US the first engineering schools opened their doors in the first half of the 19th century and by the end of the century were a common curriculum on land grant universities. By the 1930s engineers had shifted from self-employed artisans to employees working for firms. To put a fine point on it: as the 20 year old US engineering student of the 1930s entered the marriage market he brought with him a future wage superior to the average wage of his non-engineering classmate. Let’s make this guy an autistic, shave off five years and move him back a century. What are the reproductive chances of our 15 year old autistic male in 1825 America? His socio economic status (SES) will have a substantial impact. Let’s give him a meidum to high SES. He is in high school he is being introduced to “the gentler sex” at formal dances and courting is highly ritualized and monitored. I think you will agree that he is in trouble. What if he has a low SES? My guess is that with even fewer rules he is in even more trouble. The mid to high SES young man might have access to a book on etiquette, which blissfully codifies social rules. The low SES kid has to try to figure it out, and we know that this goes poorly for folks “on the spectrum.”
Let’s take this same kid back up a century to 1925. The mid to high SES version, James, has the opportunity to attend the state’s land grant college and enroll in math, biology, chemistry, engineering or physics. If the low SES student has a teacher who is willing to “stir the pot” and take the grief that will come from recommending that Johnny apply to State, and Johnny gets his parents approval, like James he too could dramatically change his reproductive prospects. This difference between James and Johnny is why the decline in communication and transportation costs is important, though to fully see why we will have to shift our scale from two 15 year old boys in a given state in the US to the globe.
I hope you have found the sexist presentation above irritating, and been wondering: what about women? I have not ignored them just because the prevalence of autism is lower among women (it is ~4:1 male to female). I have ignored them because society is sexist and the rise in reproductive success of males with autism occurred much earlier than it did for autistic women (whose social penalty is much greater since “autistic tendencies” are much more “acceptable” for a male than a female). Hays Golden’s paper sheds remarkable light on the change in autistic women’s prospects.
Golden’s paper has formal models of inheriting a gene (based on mating patterns) and income, and I do not try to summarize the details (PDF here). The short version of his mating model is this: folks partner with people like them. For his (and my) purposes, the closer you are to being “on the spectrum,” the more likely you are to desire a mate who is also autistic. He has a lot going on in the paper (e.g., shows that math skills in the US have been associated with higher income since the late 1950s), but the primary argument and evidence on which I wish to focus is the increased marriage success of women in the US with high math scores. I absolutely love it. Check this out from p. 39:
Women experienced a steep rise in the chance of graduating college, and an even steeper rise in the chance of getting a science degree. For example, 1.1% of women above age 70 got a hard science degree, while 6% of women between ages 30 and 35 got a hard science degree.
Over time has the likelihood that a woman who earned a “hard science” (yes, I do not like that term, but its a quibble) degree marries a man with a similar degree risen? You bet!
It gets even better. Autistics are rare, and it therefore follows that the smaller your pool of mates, the lower are you prospects of finding a match. We are beginning to see the drop in communication and transportation playing a role, though Golden does not explicitly mention them. Instead, he notes that sexism deterred many women from displaying and/or pursuing their ability, and that these women would pursue degrees in fields where their math ability would not remain hidden. Yet, in the partner market, both males and females would select on this hidden ability. How might one get a long run look at relevant evidence to investigate this argument?
To get at this question, we can look at the tendency of people to marry someone born in their birth state. This is a marriage outcome that depends on human capital, and is something that we can observe consistently over time (p. 41).
Now let’s have a look at a decomposition of this by the the woman’s education level.
To oversimplify but summarize, let’s take Linda and Mary, two autistic baby boomers born in Connecticut in 1947. Linda’s family has the SES to send her to boarding school, and afterward she enrolls at one of the seven sisters (none of which are in Connecticut). Mary grows up in a low SES household and drops out of high school. Not only are Linda and Mary’s reproductive prospects different, but Linda is also much more likely than Mary to produce a child with a male “on the spectrum.” And Golden’s graph of women marrying out of state shows that as the cost of communication and transportation decline—which is to say, a woman’s opportunity to attend a university outside of her state rises—more women marry someone from out of state.
This discussion of the US suggests that not only have male autistics prospects of reproductive success risen since the dawn of the 20th century, but since mid-20th century the reproductive success of female autistics has also risen, and that they chances that one of the four male autistics finds a female autistic (and vice versa) has risen as well.
But I have yet to discuss a huge intervention: the development of the integrated circuit / semiconductor / silicon chip and the consequent rise in computing power. This occurred in the late 1950s early 1960s, and by the 1970s had begun to dramatically change the earning potential of engineers and mathematicians. I had a front row seat as my father, a graduate of Virginia Tech with a degree in industrial engineering, was hired as a sales engineer in 1972 or 73 by a small silicon valley firm called Measurex. Several years later a company called Apple opened its doors across the street from Measurex, and for the next several years a number of software engineers would bounce back and forth across the street as they became irritated with working for one firm and trade it in for the other.
In any case, the subsequent rise in relative incomes for people who can write computer code has been non-linear since the 1970s, and this helps explain why we should see a considerable non-linear rise in the proportion of autistic human beings. Further, we should expect those people to exist (both through industry selection and birth) in clusters where such industries (and the universities that produce the skilled labor) exist. Both non-systematic and systematic evidence suggests that Silicon Valley, Route 28 in Boston, Austin TX, the Research Triangle in NC, and other areas with a high concentration of software (and other) engineers and mathematicians have considerably higher autism incidence than other areas (see one systematic study here).
OK, so let’s take this global. First, let’s note the drop in airfares. This graphs covers only 1980 forward, but imagine what it would look like if we could take it back to 1960 or 1950!
How about communication costs? Well, note that the populations of China and India form roughly 2/3rds of our species numbers, so drop the OECD focus and think about the modal human being on the planet. The diffusion of first radio, then cellular telephones and most recently the Internet has produced a dramatic decline in the cost of getting information about (both national and international) universities during the 20th century, and the rate of decline in that cost has been especially steep during the past two decades.
It seems reasonable to conjecture that throughout most of the planet (recall) the mid-20th century prospects of reproductive success for an autistic were not terribly different than they were for our low SES male in 1815 USA. As we move forward from 1950 to today, however, the dramatic drop in both the ability to learn about universities and the ability to travel to attend one (both nationally and internationally), should have the same impact that the diffusion of engineering programs had for American males “on the spectrum,” and access to higher education has had for American females autistics. In other words, an autistic 15 year who was born in a rural area today has much better reproductive prospects than ever before. We should be able to visit the undergraduate programs of universities throughout the world and find rurally raised autistic kids earning degrees in biology, chemistry, economics, engineering, math, physics, and statistics. In other words, though the reproductive prospects for most autistics undoubtedly remain below the mean for neurotypicals, they are nonetheless rising dramatically relative to the past.
So, there you have it: the prevalence of autism among human beings is growing, and it will continue to grow as long as (1) the cost of communication and travel continue to drop or (2) the relative wages of people with mathematics and programming skill continue to rise. Or maybe its just vaccines. F-ing neurotypicals.
PS: If you are curious about the names I use above, they are the top two most popular names in the US for that year, according to a website I queried.