Question: Which Preindustrial Society was the Most Energy-Rich?

Peter Turchin


Join 36.9K other subscribers

Some years ago I had a discussion with Ian Morris about the approach he took to quantify the social development of East versus West in his book, The Measure of Civilization. So I asked him: Which pre-industrial society was the richest in terms of energy use per capita? I have an answer to this question, which could be quite controversial (and when I offered it to Ian, I had a feeling that I didn’t persuade him).

So what’s your answer? (I have also posed this question on my Twitter)

Let’s make this question precise, so that we all use the same units. We want to measure energy use per time per capita.

Energy is measured in joules and calories (and some other more esoteric units). One calorie is the amount of energy needed to raise the temperature of 1 cubic centimeter of water by 1 degree Celsius (centigrade). 1 calorie is roughly 4.2 joules. Joules are a better unit for energy, compared to calories, because there is a confusion between 1 calorie and 1 kilo-calorie = 1000 calories. But here are the basics (taken from Box 1.3 of Vaclav Smil’s Energy and Civilization). A moderately active adult spends between 2 and 2.7 Mcal (1 million calories) per day which is roughly 10MJ (10 million joules) per day. The unit for measuring energy flow per time is called Watt = J/s (joules per second). The power of a human body, thus, works out to be roughly 100 Watts. Here’s the calculation: 10,000,000 J/(24 hours x 3600 seconds) = 115 W.

So let’s take this number as the base. In a foraging population the main energy use is the human body burning food, but let’s not forget that additional energy is needed to cook food on campfire. Smil (Boxes 1.4 and 2.1) estimates that 1 kg of dry wood contains about 20 MJ and cooking requires less than 0.5 kg of wood per day. This works out to roughly another 100 W. We have just doubled human energy use!

Bushmen getting ready to cook a meal. Source

By 1500 CE various human societies around the globe would be using a number of additional energy sources:

  1. Burning fuel for cooking and heating houses
  2. Plowing with animal power (oxen and horses)
  3. Transportation, using animal power and wind power (sailing)
  4. Energy-demanding industries: metallurgy, pottery, glass-blowing
  5. Wind and water mills to mill grain, pump water, etc.

Anything else I am missing?

Now the trick is to convert all those energy-using activities so that we can express them in per capita terms. For example, let’s do a quick calculation of how much iron metallurgy would add to energy use per capita. A peasant needs a steel axe. Let’s say its head weighs 1 kg and needs to be replaced every 5 years. Consulting Box 1.8 in Smil’s book, we find that smelting iron from ore requires 12-20 MJ/kg, and converting it to steel needs further 20-25 MJ. Let’s round it up to 50 MJ per axe head (to account for iron losses during forging). Replacing an axe every 5 years, then, would require 50 MJ/(5 years x 365 days x 24 hours x 3600 seconds) = 0.3 Watts. Well, this doesn’t seem to add a lot (assuming I did the math right).

So here’s the challenge. It’s not enough to name a particularly advanced society. Give me some numbers to show that its energy use was high.

Notify of
Most Voted
Newest Oldest
Inline Feedbacks
View all comments

Well, I read the Twitter thread, so cheating a bit.

My guess is herders who are constantly engaging in warfare and getting killed off. That would keep the per capita number pretty high.

Small populations that control vast amounts of fuel (forests/peat/coal/whatever) have a lot of potential energy, though they have to be converted (usually with human muscle).

Also, yes, maybe civilizations who underwent a recent agricultural revolution but hadn’t had the population explosion to eat away the per capita number.


Sorry. I’m stuck on ” social development”.
First choose and arbitrary term, then arbitrary criteria for that term then offer no measurements?
Its a lot of fun though.


Seriously though, I want social analysis to take neurobiology into account. There are real mechanisms in our heads with evolutionary origins. These need to become the basis of analysis going forward. Tribal Identity of tribes of 40, Empathy and lack of, Trust at a distance. These all have real biology behind them.
Heres a criteria for so ial advancememt:
What societies are structured so that tribal identity is not lost. Ie, which ones allow enough freedom that individuals can find identity and yet not seek destruction of other tribes… within their nation state.

Joel Sammallahti

I wonder about those Native American societies that would periodically burn enormous areas to promote new growth conducive to hunting & gathering.


Heh. Good point. It really depends on what you consider “energy use”. Those Native Americans got some usage out of those fires, but way less than the energy of those fires.


I was just about to write the same thing. Australian aborigines used to burn a significant portion of the continent every year.


It wasn’t always once a year. For example, nomadic peoples of arid interior Australia had a custom of setting the grass on fire wherever they went. The columns of smoke rising every few hundred meters allowed them to see where each group was going.


Mongol empire ? With, let’s say, roughly one horse for every human ?

Steve Bannister

Sung (Song) China. I will work up some numbers today or tomorrow for support. Part of my dissertation. The (Northern) Sung Dynasty CE 960–1279 essentially completed the first half of an industrial revolution, the necessary energy revolution replacing wood with coal for heating applications. They did not complete the second half of an industrial revolution, replacing muscle power with steam power (and other forms in modern industrial revolutions).


This was my answer too–I started doing some back of the envelope calculations on Twitter.

Even hunter-gatherers with energy-dense diets would use less energy per capita than a proto-industrial society like that of the Northern Song.

Steve Bannister

I have been pulling numbers from many sources in between teaching and other zooming. I have an estimate that I really need to check further because it is so surprisingly high. I think what I will do later today is create a google sheet and share it here with my step-by-step calcs, sources, and comments. Or are we out of time on this project?

Steve Bannister

Okay. Here is a link to my sheet, still a work-in-progress. It is commentable on the sheet should you wish. Topline (so far): annual per-capita energy consumption in Sung China was 70,000,000 gigajoules (GJ). Note bene this is based only on energy consumed in the production of coal, so is likely significantly understated. With sufficient interest/time, I’ll continue adding major energy uses. There is a conversion in the sheet to per-capita megatonnes of oil equivalent consumption (MTOE). Those are the units on a comparative table I will also publish as soon as I find the code and update for the Sung data.

Steve Bannister

production of iron, not coal!

Felix Aurioles

There seems to be a small mistake in the numbers. Your source puts the energy of a ton of coal between 20 and 35Gj. Or 2.3*E^10J. In your spreadsheet you put 2.3*E^10 GJ. So, once accounting for population it should be 7*E^7J, not 7*E^7GJ, once divided by (365*24*3600) It’s about 2.1-3.5 watts. Please disregard my other comment I can’t erase it.

Roy Stogner

Your math is right, but are we sure your input numbers are directly applicable?

Coincidentally I was just reading historian Bret Devereaux’s series on iron smelting:

And his estimate there is around 100kg of wood for each 1kg of finished iron equipment, a factor of 50 higher than your number. He’s measuring wet wood, not dry, so call it a factor of 40 apples-to-apples? Having to convert wood to charcoal first (at rates much less efficient than for modern charcoal production), then having to convert ore to finished product in multiple separate steps, seem to be the biggest important factors here, although the extra inefficiency of premodern smelting (where the highest temperatures couldn’t even melt the iron itself, just eventually melt the slag and get the bloom hot enough to be worked) is probably a significant difference too.

Roy Stogner

Agreed about quantification. Too much pontificating about history is still in the not-even-wrong stage, from which even being quantifiably wrong is a huge improvement.

Probably no need to correct Smil, just to think outside the box. Literally, I mean, in the text above and below Box 1.8:

“Perhaps most notably, the coke-fueled smelting of pig iron in large blast furnaces now requires less than 10% of energy per unit mass of hot metal than did the preindustrial charcoal-based production of pig iron (Smil 2016).”

So there’s at least a factor of ten difference, right there. Also, the wording “charcoal-based” seems ambiguous as to whether it’s counting energy use from charcoal alone or whether it’s also counting the energy required to produce that charcoal. If it’s the latter, then also multiply by that conversion factor (mass ratio ~7 divided by energy density ratio ~3 for wet wood = ~2.3 MJ wood used to produce 1 MJ charcoal) and Smil’s numbers end up reasonably close to Devereaux’s.

Al Kotov

all new branches of evolution came from Africa, so let’s climate that 1/f out & it feels like it

William Harshaw

You don’t consider agriculture as using energy to grow plants?

E. N. Anderson

It really depends on how you count the burning of vegetation to reshape the landscape, drive game, open land for agriculture, etc. In the long (or even medium) run it is energy-neutral, because the vegetation grows back, restoring the calories stored as wood. But in the short run it dwarfs my modern American use (car, A/C…). Looking at net consumption over time, Song China is certainly up there, as Steve Bannister says, but later China isn’t that much different.

Andrew MacLean-Finney

How about vikings living on Iceland and making use of the geothermal energy for cooking, bathing and washing.

Access to large amounts of hot water was not an easy thing in the pre-industrial era, no?


I am inclined to propose Roman society at its apogee. Why? It marshalled the energy resources and prime movers required to build a large and lasting infrastructure, feed a growing population and project its might over most of Europe. Surplus energy devoted the fabrication of boats, charriots, weapons and machines, to sustain agricultural production, feed horses and draft animals, slaves and citizens, and to produce iron, copper and other base metals. I have not done the maths yet, but directionnally, Roman society, or Rome anyway, on the availability of much surplus energy, MJ/year/capita.


Han China matched the Roman Empire in surplus energy while later major Chinese dynasties surpassed.

But they also had relatively huge populations for their time. A lot of surplus energy divided over a large denominator may still mean a lower per capita number compared to nomadic marauding herders.


Rome exported glass as far as China. Energy use for glass is not much less than for iron.


Hmm, a point that was brought up in your Twitter thread (though I thought of it as well):

Efficiency isn’t taken in to account.
OK, when it comes to food, calories/joules are roughly what they are (at least when converted by human metabolism).

But societies can be very or not very energy efficient (one society producing better steel/buildings/whatever than another at 20% the energy expenditure). The hunter-gatherers burning the land would likely not be terribly efficient (they’d reap only a small portion of the benefits of the fire, though granted, they didn’t actually have to expend that much personal energy either).

One argument in favor of the nomadic herders winning the per capita argument is that they seem to be able to bring a lot of firepower to a fight (or retreat easily to escape) vs. agricultural societies despite their much smaller numbers.

Han Wudi and descendents expended a huge amount of Han China’s output to wipe out the Xiongnu and were able to do so only by bringing the resources of massively greater numbers to bear.


Apologies for sniping from the sidelines but I would like to explore beyond the question of energy expenditure, towards quantifying human needs met, or wellbeing. Energy use is a fantastic proxy now with global industrial civilisation, but like with GDP, you can hide so much spinning of the wheels inside it. I’d like to see something that takes into account when energy isn’t used because there’s no need. When you don’t have to heat because the weather’s already warm, when you don’t have to cook as much because you can eat raw fruit, when hot water springs out of the ground and so on.

steven t johnson

Seems to me that a true comparison would be annual energy productionXman-years of productive population per areaXtotal population. And the most information is derived from a time series. .

This replaces energy flow per capita with density of energy multiplied by productivity of labor.

Obviously the statistics are even more nefarious, but this series I think is what is required. A single metric like energy flow per capita founders on the difference between nomads and civilization. (“Civilization used here in its only valid meaning, a society that lives in cities.)

Loren Petrich

“Settled societies” is a more self-explanatory term. Nomads and settled-society people have been battling it out for millennia, with the settled-society people finally winning over the last few centuries.

Ibn Khaldun identified a cycle where nomads conquer settled people, then become fat and lazy and in turn conquered by some more nomads.

steven t johnson

“Settled societies” is pretty good. The ambiguous cases where settlements are abandoned after years of occupation, to move to fresher soils, or where the group shuttles between the same summer and winter pastures are not so numerous. Distinctions within settled societies can be added as needed. Settled societies where the “cities” are basically agglomerations of clan/tribal access to a market,/.religious cult center whose population has to be sustained by in-migration of the landless differ significantly I think from those where the city wards are focused around types of good produced/sold. In particular, those settled societies where urban populations can reproduce themselves are functionally different from most societies.

Edward Turner

The correct unit of analysis for this exercise is the polit because civilisation is thing of culture that belongs to and is produced by whole societies, not individuals.

So we need to calculate the energy the polity used to survive whilst being and producing this civilisation.

As hard as it may be to calculate, this includes the energy used to cut, fashion and transport materials and goods from A to B, which could include human, pack animal, wind energy, or the current energy of a river, when goods are transported downstream.

It includes solar energy for the production of man-made agriculture like grains and products such as dried fruits which enrich the society with greater variety, which is a characteristic of a more sophisticated civilisation.

Essentially a great civilisation harnesses energy sources much more effectively than civilisations that are not so great.

We should divide the energy figure per capita to compare different polities/civilizations but I am not interested in the calorific consumption of individuals specifically for this measure because the unit of analysis for consumption of energy is the polity because it’s the polity that produces the culture that creates the civilisation.


How about good, old-fashioned agriculture? Specifically for a society that had lots of solar energy, and a low population because of other constraints, like water. I’ll do an analysis for the Hopi Tribe of the American southwest, although fair warning: a lot of these numbers are just guesses.

Farmland per person: 1-2 acres, according to this:

Average solar energy per sq meter: > 6.5 KwH per day, From this:

1 acre * 4000 m^2/acre * 6.5 KwH per day / 24 hours per day = 1,083,333 watts per person! or 1444 horsepower!

Of course, not all of that energy is being “used”. Most of it just bounces away as radiant heat (but that’s a problem with most of the other calculations you’re doing, too). According to this: a modern cornfield converts about 1.5% of the suns energy into edible food, so even if they were just 1% efficient, that’s still over 10,000 watts per person.

Felix Aurioles

I apologize for the late reply.
When I first read “why the west rules for now”, and “the measure of civilization” I was thinking the same thing: That there is no nomad paradox, as you said, nomads are incredibly wealthy when compared to settled pre industrial peoples.
Similarly all that moving around, produces and transmits a constant stream of information about prices, environmental conditions, social situations, risk calculations, etc.
The only thing left unexplained is the largest settlement part, but one can very well argue the size of the largest settlement, is but a proxy for the logistic capacity of a given society. Largest gathering would probably be a more general measure, or alternatively largest gathering lasting over a certain time threshold. That way we could include the moving of a quarter million men, their horses, equipment, and so on during a campaign.
One could also argue that Nomads don’t exist by themselves, and nomadic empires evolve along with urban centers.

As for the output calculations.
According to Smil’s energy and civilization a horse can produce 11Mj of work a day, which is about 127W. Smil also suggests pastoralist society would require, at a minimum, 2.5 horses (or camels) per capita. When adding the requirements for cooking and food, we will be dealing with about 517W per person. It should actually be a little more than that, for we are not counting the energy needed to produce the commodities used.

So Pastoralists 517W at a minimum

Ian Morris seems to count the man hours of labor that went into the manufacture of a given commodity. So a complete estimate would involve the fashioning of an average basket of goods a society consumes, and the amount of energy required to produce each one, including human labor. If we are wiggling to assume urban dwellers do not participate in the production of food, or if they do it is to the same extent that rural dwellers participate in the manufacture of commodities.
Total energy used = the caloric consumption, + (percentage of urban population)*constant relating human work output + other sources of energy.
As to be fair to settled cultures, let’s grab the Netherlands just on the eve of industrialization, also because most of the information we need seems to be neatly arranged in Jan de Vries’s “first modern economy”:

Total energy used = the caloric consumption, + (percentage of urban population)*constant relating human work output + Energy employed in shipping/population + windmills/population + other stuff.

For caloric consumption we use the same base value as in the OP, 200w

42% of its population is urban in 1750. (Page 83 Vries’s “first modern economy”) Keep in mind the Netherlands is an outlying case. Around this same period England has about 16% of urban population. Going back to Smil, a person can deliver 2Mj of useful work per day, which comes to 23W. 42% of this is 9.7W

A term for wind transport (page 462 Vries’s “first modern economy”) a maximum of 700 million ton-miles of shipping, representing about 40% of Europe’s shipping, this is again an uncommonly high number for a settled society. 700 million *1000(kg per ton) * 1852(meters in a nautical mile)/ (365*24*3600*2.1 million (Dutchmen)). It comes to about 10.54W

Let’s add a term for the Windmills, Smil, reports a “Typical windmill” to produce 7.5KW. There were 12000 of them in 1800, 7.5*1000*12000/1.951million or 90/1.95, 46W. This seems a bit high in light of everything else, but I couldn’t find reliable numbers for 1750.

So we have 200 + 9.7+10.5+46= 266.35 W
We can round it to 270 W to account for the energy employed in manufacturing steel. Still slightly over half of what a pastoralist needs to be ecologically viable.

Felix Aurioles

I’m glad to participate. I have been enjoying your work for a while, your explanation of Fick’s law at the appendix of “Quantitative analysis of movement” is very clear and easy to follow, more so even than many textbooks specially dedicated to diffusion/transport phenomena.

As to why I did the accounting that way; Ian Morris uses energy consumed to approximate a standard of living measure. Which makes sense in light of the correlation between GDP and energy spent across much of history. So I tried to do a similar Accounting.
From “measure of civilization”:

“Classical Greek houses were large and comfortable, typically having 240-320 m2 of roofed space. The evidence for house prices is disputed,47 but an average house probably cost 1,500-3,000 drach­mas at a time when a 5,000 kcal daily diet cost about half a drachma? Meaning that an average house represented 15-30 million kcal. Amortized out over a thirty-year lifespan, that represented close to 1,375-2,750 kcal/day. (There is no way to know what Greek expec­tations about the lifetime of a house were, but thirty years seems roughly consistent with the rate of rebuilding observed on archaeo­logical sites”

And later:

“It is also striking that classical Greece supported not just relatively high levels of nonfood consumption but also high population densities around the Aegean Sea in the fourth century BCe. In several parts of Greece, the densities of the fourth century BCe would not be equaled until the twentieth century Ce, and the simple fact that so many Greeks lived in towns or small cities, rather than hamlets or farms, must mean that their energy capture reached unusual heights.”

He however accounts the food needed for the laborer rather than the work output. I didn’t do this because, since it is a per capita measurement it counts the food twice, once when he eats, and then again as his labor output. In that particular example, it is also only one person living in the house inflating energy use. But that is beside the point.

However, one would have to count the horse energy as it relates to feeding stock, so pastoralist again would be using a proportionally larger amount of energy. In the Holland example, even if the energy proportion relating to urban activity would be increased by an order of magnitude when using Morris’s accounting, the terms relating to Windmills and sailboat would remain the same and be a smaller proportion of the total. Which does not agree with Smil’s data of a small windmill doing the work of a dozen men.

As to transportation, I imagine a significant portion of the energy available to pastoralists is spent moving things around, their camps, their trade goods, and their herds to other areas. Since merchants are usually wealthy in comparison to other members of their societies, arbitrage would seem to be a significant portion of economic activity. If we expect the 4 indexes Morris uses to have some correlation, the index pertaining to information production and exchange would correspond to a movement/trade term when accounting for energy. Since a lot of the information used by human societies is in the form of prices.

Felix Aurioles

Thank you. It is a very good reference book.

I found this list:
which seems to produce database:

So apparently, following those criteria. The wind mills and sailboats would be counted but not the horses, camels or oxen. The wood used for cooking, but not the food itself. Human labor, either in the way Morris, or as I estimated, would not contribute to the total.

It probably makes sense to do it that way for industrial societies, after all, muscle labor represents a small fraction of the energy spent. But as the previous estimates in this thread have shown, during much of history it was the opposite; so another measure may be more appropriate. I think the methods of accounting we use can illustrate slightly different processes. And so, once we have a more defined model of energy flows we could probably polish our measures.

I came up with some objections and advantages to different accounting methods.
Thinking back on the accounting Dr. Morris uses, it assumes the monetary value of energy remains constant independently of the form it takes. This unfortunately results in the energy embed in a house to be calculated in a food-price equivalent, when the energy involved was probably human or animal labor. But thermodynamic constraints on metabolism, ensure the energy output of an animal would be an order of magnitude or so smaller than its caloric intake, so both types of energy should not be fungible. Which is what I believed his count implied that was the total energy including food needed to produce the work used by the construction.

So if the energy of human/animal labor is not fungible to that of other sources. Maybe the solution is to count energy types according to their “quality”, a property analogous to thermodynamic Exergy. One count for diet, one for fuel, one for human work, one for animal work, etc. See how they relate to other state variables in the society. And then estimate the not-exergy as a function of how that relationship changes over time.

Another approach would be to treat calories consumed as an output, and account the labor inputs required to produce them, but we would just end up counting every person and every animal. It would look something like this:

Energy used per capita = a constant + the number of work animals per capita + other types of energy (coal, wind, etc).

A solution would be to do away with the work needed for agriculture altogether, since the output is implied in the caloric constant. If we can assume other energy inputs into agriculture reduce the need human muscle, then the share of population not dedicated to agriculture would grow, as would the total energy use.

This last bit reminds me of Dr. Colinvaux’s “the fates of nations”: where he suggests the depopulation of the Roman countryside in the imperial era and the formation of latifundia was not indicative of “decay”, but of rising productivity displacing farmers and creating a larger fraction of urban populations. In your own, “Secular Cycles”, you show an (slight) inverse relationship between agricultural yields and real wages and in some cases an increase of urbanization just before the collapse. So this accounting method could tie in energy use with historical dynamics.

Hopefully I am not taking too much of your time with these long replies.

Felix Aurioles

I am glad we agree. I will try to write the argument in more formal terms, in preparation for our next exchange.

steven t johnson

In rugged terrain—which I understand most of Greece is—population tends to cluster more in order to leave the arable (flat) land available for use. Access to water tends to be at open sources like rivers rather than wells, which leads to greater clustering. Not sure if hilly terrain gives better access to stone and timber but that seems plausible at first glance. Lastly, in the specific Greek case, clustering at access to cheap water transportation seems likely to be a structural bias as well. This last may be a factor in the Dutch/Belgian case.

The question of how much these factors contributed is the point, though.


Energy use is poorly defined. Arguably, a farmer uses solar energy to grow crops. Does this count? If not why not?

steven t johnson

If horses didn’t eat solar energy captured by grass, the point would be inescapable. As is, counting solar energy at the plant level is more like double counting I think. Also, forage crops confuse this issue too, as does animal power for traction.


Energy use is a possible measure, but a lot depends on what one counts as a civilized life. If you live in a cold climate, you explicitly need energy for heating. If you are living in a warm climate, you can get that energy from solar power. If you live in a hot climate, odds are you would like to or currently are expending energy to stay cooler.

There’s also the city problem. The whole point of cities is to cut energy use. You didn’t need your own cooking fire, you used the lord’s or the baker’s. You had a thermocline for warmth in the winter. You had proximity to cut transportation energy use. Even now, people living in cities in the developed world use less energy than their rural counterparts.

That brings up the issue of energy conservation. If we replace incandescent bulbs with LEDs or improve the fuel efficiency of SUVs ,does that mean we are becoming a less advanced civilization? Smil is oriented towards peak energy use as a technological indicator, so high pressure steam beats low pressure steam, horses beat oxen, fusion beats fission. Moving from working with bronze and pottery to working with iron and glass is a major step, but what happens if we replace iron and glass with biologically engineered hydrocarbons?

  1. Home
  2. /
  3. Cliodynamica
  4. /
  5. Regular Posts
  6. /
  7. Question: Which Preindustrial Society...

© Peter Turchin 2023 All rights reserved

Privacy Policy