Published 07 February, 2020; last updated 28 May, 2020
The number of books produced in the previous hundred years, sampled every hundred or fifty years between 600AD to 1800AD contains five greater than 10-year discontinuities, four of them greater than 100 years. The last two follow the invention of the printing press in 1492.
The real price of books dropped precipitously following the invention of the printing press, but the longer term trend is sufficiently ambiguous that this may not represent a substantial discontinuity.
The rate of progress of book production changed shortly after the invention of the printing press, from a doubling time of 104 years to 43 years.
This case study is part of AI Impacts’ discontinuous progress investigation.
Around 1439, Johannes Gutenburg invented a machine for making books commonly referred to as “the printing press”. The printing press was used to quickly copy pre-created sheets of letters of ink onto a print medium.1 Presses that stamped paper with carved blocks of wood covered in ink were already being used in Europe,2 but Gutenburg made several major improvements on existing methods, notably creating the hand mould, a device which allowed for quickly creating sheets of inked letters rather than carving them out of wood. The printing press allowed for the quick and cheap production of printed books like never before.3
We looked primarily at two different metrics– the rate of book production in Western Europe and the real price of books in England. We chose these two because they were some of the only printing-related data sources which had data that went back several centuries before the invention of the printing press.
Had the data been available, we would like to have looked at some metric correlated clearly with innovations in the writing / printing process — e.g. the number of pages produced per worker per hour. Then we could check whether the printing press represented a discontinuity relative to earlier innovations (e.g., the pecia system for hand-copying manuscripts).5
Unfortunately, neither the rate of book production nor the price of book data we have correlate well with innovations in the writing / printing process. The authors of our rate of book production data claim that most of the variation in the pre-printing press numbers is explained by factors which are not innovation or close proxies to innovation.6 Our data on the price of books is similarly unhelpful, as the early price data is too sparse to be meaningful.
In addition to the two metrics described above, we looked cursorily at a few metrics with no early data which changed drastically as a result of the printing press– the number of unique titles printed per year, the variation of genres in books, the price of books in the Netherlands, and the total consumption of books.
Our data for the rate of book production come from estimates of Europe-only production generated in a 2009 paper by historians Eltjo Buringh and Jan Luiten Van Zanden.7 Rate data is represented as the number of books produced in the previous 100 years at various points in time.
When we use the term book, we mean it to refer to any copy of a written work, whether hand-copied manually or produced via some kind of printing technique. The paper separates book production into estimates of manuscript and printed book production, where the production of printed books starts only after the printing press is invented. We will also use the terms manuscript and printed book to talk about the data, but it’s unclear to us if the paper means manuscript to mean “any book not made using a Gutenburg-era printing press” or “any book transcribed by hand”. At one point the authors sum these two estimates into a single graph of production per capita,8 suggesting that the combination of manuscript and printed book data should cover all books.
The paper’s estimates for manuscript production are constructed by taking an existing sample of manuscripts and then attempting to correct for its geographical and temporal biases.9 Estimates for book production are constructed by counting new titles in library catalogues and multiplying by estimates for average prints per title at a given time.10
The estimates of manuscript production seem extremely non-robust given that large number of correction factors applied.11 The estimates of book production seem somewhat more robust, but should be taken as a lower bound as the authors did not correct for lost books and have estimated the average number of prints per title conservatively.12
Figure 1a displays the raw data for rate of book production on a log scale, taken from the data in the paper described above and compiled in this spreadsheet. Each data point represents the total number of books produced in the previous 100 years.
Figure 1b displays the same data as Figure 1a along with our interpretation.
Looking at the data, we assume an exponential trend up until 1500, and another one after that.13 The blue line is the average log rate of the rate of book production before the invention of the printing press (just manuscripts); the red line is the average log rate of the rate of book production after the invention of the printing press (manuscripts + printed books).
Grey points shown after 1500 reflect projected manuscript and therefore book production had the printing press not been invented. In practice, the actual number of manuscripts produced after 1500 were very small and not presented in the data.
If we just look at the trend of book production per past 100 years, measured once every 100 years before 1500, and then once every 50 years afterwards, we can calculate discontinuities of sizes 161 years in 900, 134 years in 1200, 23 years in 1300, 180 years in 1500, and 138 years in 1550.14 This is obviously a strange kind of trend–a discontinuity of one hundred years in a metric with datapoints every hundred years might mean nothing perceptible at the one-year scale. So in particular, these discontinuities do not tell us much about whether there would be discontinuities in a more natural metric, such as annual book production.
There was a marked change in progress in the rate of book production with the invention of the printing press, corresponding to a change in the doubling time of the rate of book production from 104 years to 43 years.15
Interpreting this rate of change on the graph, before the invention of the printing press, the total rate of book production, which consists entirely of manuscripts, follows the exponential line shown in blue. The invention of the printing press in 1439 allows for mass production of printed books, causing the rate of book production to veer sharply off the existing exponential line, shown as the first point in red. Note that our underlying data sources are non-robust, particularly for manuscript data pre-printing press, so the magnitude of this change in rate of progress may be under or overstated.
The change in the doubling time of the rate of book production caused by the printing press may reflect a large change in the factors that drove book production.
In their paper, Buringh and Van Zanden note that in the Middle Ages, 60% of the variation in book production is explained by the number of universities, the number of monasteries, and urbanization.16 They produce the following graph, correlating monastery numbers and early book production:
By contrast, after the printing press was invented, Buringh and Van Zanden attribute a much more important role to individual book consumption and the forces of the market:
How to explain the significant increase in book production and consumption in the centuries following the invention of moveable type printing in the 1450s? The effect of the new technology (and import technological changes in the production of paper) was that from the 1470s on, book prices declined very rapidly. This had number of effects: consumption per literate individual increased, but it also became more desirable and less costly to become literate. Moreover, economies of scale in the printing industry led to further price reductions stimulating even more growth in book consumption.
It seems plausible that the move between exponential curves caused by the printing press was a shift from an exponential curve that reflected the growth of monasteries, universities, and cities to an exponential curve that reflected the growth of a complicated set of market forces.
We took data from a 2004 paper written by economic historian Gregory Clark, who took data from historic records of price quotes,17 though data before 1450 is based on just 32 total price quotes, and Clark notes that “the prices vary a lot by decade since it is hard to control for the quality and size of the manuscript.”18 As such, we should not take much meaning out of the individual data points before 1450 or interpret the period between 1360 and 1500 as a rising trend.
Clark’s paper reports an index of the nominal price of books in Table 9 of his paper. To instead get an index of the real price of books,19 we divided each nominal price by Clark’s reported “cost of living” for each year (x 100), which was the amount of money paid by a relatively prosperous consumer for the same bundle of goods every year.20
Figure 3 is a graph of (an index of) the real price of books in England, generated from the data described above and compiled in this spreadsheet. Each point represents the amount of money in each year needed to buy some bundle of books assuming you would pay 100 for that same bundle of books in 1860.
Economist Timothy Irwin, looking at this same dataset, claims that the drop in price around 1350 was due to knowledge about paper-making finally making its way to England: “As time passed, improvements in technology and the gradual spread of literacy reduced these obstacles to effective transparency. In particular, the diffusion of knowledge about paper-making (from around 1150) and then printing (from around 1450) dramatically reduced the price of books. (See Figure 1 for an estimate of the decline in England).”21
‘Figure 1’ in the quote above refers to a graph of the real price of books that uses the same data source and is identical to the one we generated. His claim seems plausible, but we are not confident about it given how sparse and noisy the early data is and given our lack of precise date information on how paper-making spread in England.
The graph contains two major drops– one as a result of the printing press, and one claimed to be the result of paper replacing parchment in England. Looking at just the set of blue data points between these two drops, we can see that they are clustered around some range of values. The data in this range is too noisy and sparse22 to generate a meaningful rate of progress, but it seems clear that for a wide variety of plausible rates, the printing press represented a discontinuity in the real price of books when compared to the trend in book prices after the spread of paper-making.
However, if you take the past trend to include the price of books before paper-making, then there is no clear discontinuity– the price of printed books could be part of an existing trend of dropping prices that started with paper-making. We also believe the data here is too poor to draw firm conclusions.
Whether or not it counts as a substantial discontinuity relative to the longer term trend, the printing press produced a sharp drop in the real price of books. This was because their price was largely driven by labor costs, which went down sharply (one author estimates by a factor of 341) when a laborer could use a machine to print massive numbers of books rather than manually transcribing each copy.23
Many historians associate the invention of the printing press with other unsurprising book-related changes in the world, including:24
Most of these changes are gradual over at least a century, rather than involving a sharp change that might be a large discontinuity. Such gradual changes might suggest a sharper change in some underlying technology though, for instance. The fall in the price of Dutch books was relatively abrupt, but the data lacks a trend leading up to the printing press.
The increases in the number of genres and unique titles published suggest that there was a larger amount of information available in printed form. Decreased prices and increased consumption of books suggests that this information was easier to access than before. These things suggest there might have been an interesting change in availability of information in general, however we do not know enough about the past trend to say whether this was likely to be discontinuous.