How to Predict the Future

Image-empty-state.png

Social science – or at least quantitative social science – claims to be able to predict the future. Among the various criteria on which theories are evaluated is predictive power – the ability of the model to make accurate predictions. Economists and political scientists get paid large consulting fees for forecasting the economy, election outcomes and social conditions in general. Macro social scientists, such as myself, write books on when and how societies fall.


That said, a lot of social science predictions go awry. I don’t even have to mention how many election predictions look bad on Wednesday morning. Paul Krugman likes to remind everyone of how many economists said that X tax or Y government deficit was going to lead to massive inflation and slow growth. (Punch line: None of the inflation and slow growth actually occurred.)


For really painful cases of social predictions gone awry, I read the sociology of the 1950s and 1960’s. We were becoming a full employment leisure society. Everyone would have plenty of money without having to work. How to use all our free time was going to be our biggest problem. Recessions, depressions and social conflicts would be a thing of the past. The only question would be when the poor nations of the world would join us.


For the most repetitive cases of social predictions going sour , read Neo-Marxist analyses of how the most recent current-event-de-jour is leading to a crisis in capitalism. Some of these do correctly identify persistent problems that refuse to go away and lead to significant conflicts. But I can get very tired of reading “current crisis in capitalism” papers that identify problems that do not slow down capital accumulation one bit. The bankers go on making money. The workers stay home watching TV. The next day and the next decade, capitalism is still there.


It is still possible for social scientists do some predicting. It is helpful to be modest about what they can do. It is even more helpful to avoid the potholes that deep-six a lot of sociological predictions. Forecasters make the same mistakes over and over again. However, the same opportunities exist over and over again to say something useful.


Here are some of the DOs and DON’Ts.


DON’T. Don’t Always Assume That What Will Occur in the Future Is Simply a Continuation of What Happened Recently. Making social forecasts based on an extension of recent trends is known as extrapolation. If the economy of your city has been growing 3% a year for the past 5 years, you assume your city will continue growing at 3% a year forever. The no-brainer way to do an extrapolation is to make a graph of changes in what you want to study over time. Use a ruler to make all of those changes look kinda sorta like a straight line. Extend that line in the future. If in the last few years, your city grew 1%, 5%, 2.7%, 3.3% for various years …. Eh, call it 3%. With a ruler extend a line showing 3% growth from now to infinity. Congratulations, you have your prediction.


If your customer is likely to think this strategy is a little too shabby and a little too obvious, time to pull out the mathematical razzle-dazzle. Don’t just use one indicator of change, use five or ten or thirty indicators of change, combine them all using a computer algorithm and a mathematical formula your customer might have a hard time understanding. Have the computer assume all of these things stay the same forever. Have the computer generate the line. Want something a little more elaborate than a line? Arbitrarily make a few things quadratic, exponential or logarithmic so you get a curve. The choice of math function you choose arbitrarily. The computer takes the math function of the past, mindlessly extends it to the future, and you have your prediction. The future will be the same as the past, but with a more complicated algorithm.


Obviously, you can see where this whole system can go wrong. If there is any fundamental change that makes the future profoundly different than the past, this whole extension from past experience is dubious. In the 1950s social projections were made assuming that we would have access to unlimited fuel, energy and natural resources … because we had always had access to unlimited fuel, energy and natural resources. Assume that none of our growth would be harming the environment, because we hadn’t done anything really harmful yet. Assume America would be a hegemonic power with economic monopolies over nearly all technological goods, because we had been a hegemonic power with monopolies ever since 1945. The 1973 OPEC oil shock and the rise of Japanese competition came as huge, huge surprises.


A few things really can be predicted from simple extrapolation. They are caused by a small number of known factors which change very slowly over time. Population growth is one of the few phenomena that can be predicted well from extrapolation. Even dramatic events such as pandemics or war, do not have that much effect on long term population trends. If you know what happened to fertility and mortality in the last thirty years, you can extend your population estimates forward with some level of confidence.


Few other phenomena are as stable, orderly or predictable as population. Economies and governments are NEVER that stable, orderly or predictable.


DO. Do look for long term historical cycles. More things cycle than trend.  What goes up comes down. What has been down for a while goes up. History is a long run of vacillations. Economies boom and bust. Imperial powers rise and fall. People seek personal liberty – People seek moral order.


Societies vacillate because nothing works all the time. Marx used to talk about the contradictions between the forces and relations of production. The forces of production are the short term things business needs to do to remain profitable. The relations of production are the long term things business needs to do to remain profitable. You don’t need to be a Marxist to know that what businesses and economies do to make money today lays the seeds for future troubles tomorrow. Markets get saturated. Resources get exhausted. Your customer learns to how to make your product and enters into direct competition with you. The best laid plans of everybody run into problems later on.


When something stops working, people start wanting the opposite. The economy not working? Bring in a strong government to take dramatic action. The economy not growing under the dictator? Throw the bum out and let the spirit of freedom and free enterprise rise. Energy prices rising? Buy bikes and fuel-efficient cars. Got the fuel problem under control? Bring back the SUV!


You can do a lot of good predicting by looking at the current situation, and figuring out what can go wrong.


Assume it will go wrong.


Now what?


DO. Do learn the difference between structural and conjunctural processes.


DON’T. Don’t even try to predict the conjunctural stuff.


Structural processes are those that are determined by general sociological laws. They are study-able by social scientists. They create phenomena that are observable in many societies, and repeat themselves often in those societies. They create tendencies where the majority of cases follow the social laws while exceptions are more likely to be rare. The laws are confirmed by multiple studies in multiple settings that all tend to show the same things.


Examples of structural processes:


The children of rich parents tend to be richer than the children of poor parents.


Governments tend to respect the preferences of the largest enterprises in their country.


Wealthy nations attract migrants.


Sociologists are in the business of finding sociological laws. We have found a lot of them:


Rich countries become richer by obtaining the resources of poor countries.


Governments with insufficient resources become politically unstable.


Countries with more education become more feminist.


In contrast, conjunctural processes are based on fluke or chance. Often, the actions of idiosyncratic individuals play a big role here.


Examples of historical conjunctures:


A general blunders his way through a big battle; what should have been a major win has turned into a crushing loss.


Imagine a world where Hitler did not overrule his generals. Imagine Germany winning World War II.


A new religion takes hold. The hierarchs of the old religion are condemned to irrelevance and powerlessness.


Imagine the religious life of the West if Christianity had just been another gnostic cult in Palestine. Imagine nobody believing the bit about Christ rising from his tomb.


An inventor creates an amazing new product turning a backwater city into an economic powerhouse.


Imagine Henry Ford not getting his idea about an assembly line. 


Imagine Detroit being no more than a northern version of Toledo.


Social scientists who stick to structural regularities can make some pretty good predictions. Any social scientist who claims to be able to predict conjunctures should not be taken seriously.


DO. Do play the odds and predict the properties of whole categories of event


DON’T. Do not try to predict specific events.


In my own book, All Societies Die, I have extensive discussions of economic collapse. I carefully avoid saying the Economy of Nation X will fall on Year Y at Time Q Based on Problems in Industry J. I do not try to predict if individual stocks will go up or down.


There are general societal configurations that are consistent with economic growth.


There are general societal configurations that are detrimental to economic growth.


Societies that have more of the good stuff will tend to grow.

Societies that have more of the bad stuff will tend to have problems.

I don’t specify that Sweden will do this in 2025 or that the Congo will do that in 2038.


Those cases will be heavily determined by conjunctures.


But societies that have more of the structural foundations of growth will as a group be more likely to be prosperous. There will be exceptions.


Societies that have few of the structural foundations of growth are less likely to be prosperous. There will be exceptions here too.


*  *  *


Cover the general trends and you will be right more often than you will be wrong.


That is the best that a social forecaster can hope to accomplish.