Saturday, November 06, 2010

On projections and predictions and their flaws

I'm a big fan of Canadian journalist Dan Gardner's new book Future Babble: Why Expert Predictions Fail - and Why We Believe Them Anyway. Concerned with examining why so many confident people make so many predictions which tend to almost always be wrong, Gardner examines the many fields in which predictions are made. Demographics feature prominently.

"To look into the future is not always an idle fantasy or a wasteful foolishness," wrote Robert Sencourt in 1925. A nation's power is based on the robustness of its population--in demography "the parchment of fate unrolls"--and so, by looking at fertility rates, it is possible to forecast which countries would be the dominant powers of Europe in 1950. The future belonged to Italy and Spain, Sencourt concluded. In 1922, the American political scientist and journalist Lothrop Stoddard cast a wider gaze and came to a more frightening conclusion. "Ours is a solemn moment," he wrote in his widely read and influential book The Rising Tide of Color Against White World Supremacy. "There is no immediate danger of the world being swamped by black blood. But there is a very imminent danger that the white stocks may be swamped by Asiatic blood." This would mean the end of civilization, apparently.

In the 1930s, as fertility rates continued to decline, experts increasingly worried that fertility would slip below the level needed simply to replace populations. Growth would stop, then populations would shrink. A 1934 projection showed Britain's population falling from 44 million to 33 million by 1976. Fear that the West's political and economic fortunes would follow the population's trajectory became widespread. This was one of the "most critical epochs" in all of human history, wrote Aldous Huxley, author of
Brave New World, in a British magazine. "Will the depopulation of Western Europe and North America proceed to the point of extinction or military annihilation?" (48-49)

It goes without saying that these predictions did not take place. Projections even 40 years into the future--barely two generations, if that--were fundamentally wrong. Assumptions that fertility outside of North America and Western Europe would remain high indefinitely were wrong. Considerations of the effects of technology and organization and levels of economic development and education and health on human capital were barely taken into consideration, if ever. The result of these and other errors was terribly flawed statistics.

I'd like to think that, here at Demography Matters, we've been good about this. The post that comes to mind approaching these errors is this one I made concerning Eurostat metadata projections for populations and population structures in European Union member-states by 2060. The projections are indicative of the ways things are expected to proceed, at least. I think it's possible to make some medium-run projections with some amount of certainty. I expect that those countries and societies which are kindest to women, allowing for a balancing of employment and parenting responsibilities between genders and not forcing women to choose between parenting and family, which tend to postpone family formation (childbirth particularly) and where marital relations tend towards the informal union if that, will experience higher net replacement rates than those societies which don't. I've a shortlist of countries and societies where I expect populations will start declining and aging soon, and societies where the reverse will happpen. I expect that pressure for migration from relatively poor to relatively rich societies will continue, the volumes being limited by the receptiveness of receiving countries and the numbers of people of migrant age in sending countries. I expect that as people live longer and healthier lives, pressure for the reform of social security systems will grow sharply. I'm almost entirely certain that the idea of Eurabia is ludicrous, inasmuch as it overlooks things like changing and converging fertility rates and diverse streams of migrants and the weakness of a common "Muslim" identity (among many other things).

I'd like to think that these are fairly well justified, fairly conservative predictions, with details limited to fairly short timespans and predictions about the effects of cultural and economic dynamics that may continue for longer periods of time. The sort of radical doubt that Gardner identifies as key to making good predictions is something I'd like to think I've adopted for myself, and that I think my co-bloggers here have adopted. I think we've got a good track record. Certainly we're writing and arguing in good faith.

And yet, sometimes there can be radical changes, transformations which occur in very short periods of time. The case of Spain comes quickly to mind. In the 1990s, Spain's population was expected to plateau in the near future at 40 million and to start a sharp decline. Instead, Spain's economic boom led to mass immigration, from societies in Latin America and eastern Europe and North Africa, to the point of reversing population trends entirely and leading to population growth of more than a fifth over 1999-2009. This was an entirely unexpected event that, in a very short period of time, changed the basic trend in Spain. What other trends and events, I wonder, shall appear out of nowhere in the near future?


Colin Reid said...

Well, the people giving doom-laden predictions about Western demographics weren't completely wrong when it came to Europe (they were about the US, though). Europe's share of world population is a lot smaller than it was in 1900, despite immigration, and Europe's geopolitical influence has diminished as well, despite increased coordination between European countries (though given that Europeans and their descendants literally ruled most of the world in 1900, that's not saying much).

On the other hand, there are plenty of countries with much larger populations than they used to have, but not much influence because their economies are so weak. This was true even of China and India for much of the 20th century. So I think the bigger mistake these predictions made was to overestimate how quickly the rest of the world would catch up with the West economically.

Randy said...

Relative decline, sure, but the huge absolute decline predicted didn't happen. The predicted growth was the reverse of what happened--instead of shrinking by a quarter over that timeframe, Britain's population _grew_ by a quarter. Likewise, we didn't get a France of 28 million by the early 1980s. This unexpected growth had huge consequences, on everything from the relative powers of European states (Spain and Italy did not dominate) to Cold War balances of power (if the Soviet Union had much larger youth cohorts than NATO Europe, what could have happened?) to the huge post-war boom enjoyed by Europe (more difficult if there'd been rapid aging and workforce shrinkage) to all manner of cultural side-effects I can't imagine.

As for catchup, I'm not sure that those predictions were wholly right. Populations in MENA grew hugely over the 20th century, relatively and absolutely (there were 30 million French versus 3 million Algerians in 1930, versus something predicted to be closer to equality by the mid-21st cenutry), but that wasn't translated into particular heft despite decent starting points. As for India and China, it's not inconceivable that they couldn't have kept their low-growth rates; the survival of the institutions responsible for the "Hindu rate of growth" and a different outcome to post-Mao succession disputes could have kept them from growing as they have.

Anonymous said...

Speaking of Spain and predictions.

Someone on this blog predicted that by 2020 Spain would probably have a higher TFR than Morocco. I think it was you Randy. This sounds like an entirely reasonable possibility. I was just wondering if the person still holds to the prediction (I don't remember precisely if it was a full fledged prediction or not).

Anonymous said...

Sorry, I found the comment about the 2020 thing and it was from the Blogger Edward, not from Randy.

It wasn't really a prediction either. Just said that he wouldn't be suprised.

Randy said...

Those sorts of open-ended projections are probably the best sort. Probabilistic ones are best, IMHO; if you're making projections about the future, trying for excessive specificity isn't a good idea. If anything, trying for less might be better.

michaelgrayer said...

I think in common parlance, the concepts of "projections" and "predictions" have become so mixed up to the point that the words seem interchangeable. They are not.

Predictions involve statements along the lines of "I think [xyz] will come to pass in [timeframe]". These are highly speculative and convey a sense of expert opinion. Projections, on the other hand, will take a trend (such as population growth), break it down into component parameters (such as birth rates, death rates, migration), apply certain values to those parameters, and see what happens to the overall trend as a whole.

Projections are arguably more useful, and though they can be used to make predictions, they are arguably best used for other purposes, such as goal setting, by setting a target and seeing which combination of parameters will allow you to achieve that target. Projections often don't come to pass, but often their purpose is not to try to predict that.