Interactive companion

How Do Information Markets Respond to a Crisis? Plague, Print, and Persecution in Early Modern Europe

Noel D. Johnson and Alexander Taylor · Research

1Argument

A long tradition holds that epidemics breed scapegoats. The paper tests that claim on the presses of early modern Europe. Plague returned to the cities of Europe for three centuries after the Black Death and it returned to any one city at times close to random. Each return was a shock to what a town's readers wanted. The paper follows what the presses printed at 607 clean onsets in 173 printing cities and asks whether persecution followed.

The presses printed medicine. In the onset year the medical share of a city's print nearly doubled, from about 26 titles per thousand to about 49. The rise was outbreak medicine. It consisted of practical regimens in the vernacular, was printed as new works rather than reprints, and was largest where the city's own stock of such titles was thin. Print aimed at Jews and accused witches did not rise. Witch trials did not rise and persecutions of Jewish communities show no detectable rise.

The reason lies in what kind of product each kind of print was. News lost its value within months and plague medicine within about five years. Both were printed near their readers and in their language. Doctrine, the treatise literature of persecution, held its value for decades and was traded across the continent. Of the news a city's readers could reach, about one part in 55 came from the city's own presses that year. Of the doctrine, about one part in 850 did. No local shock could move a stock of that kind.

2Four products

The paper treats print as several products that differ in how fast they lose their value and how far they are traded. Table 1 of the paper sets out four.

ProductLoses its value overMarketBuyerSupply
Table 1. Four information products, from Table 1 of the paper.

A local shock should raise local printing of a product when the product is perishable and its readers are near the press. When the product is durable and traded the stock already within reach meets a local shock and local printing need not move.

3The returns of plague

Biraben's chronology records plague outbreaks town by town. The paper defines a clean onset as an outbreak recorded within 10 kilometers of a city after at least five plague-free years. The 224 cities of the print panel had 946 clean onsets. The 607 that fall in years when the city printed, and outside any other onset's window, enter the event study.

Figure 1. Upper panel, localities with a recorded plague outbreak in the selected year (rust) and the 224 cities of the print panel (navy). A ringed city had a clean onset that year. Lower panel, recorded plague outbreaks by year, 1347–1700, as in Figure 17 of the paper. Click a bar to jump to its year.

4What the presses printed

For each clean onset the paper follows the share of each kind of print from five years before to five years after, with city and year fixed effects and the year before onset as the reference. Every band in this section uses the wider of the city-clustered and Conley standard errors, as the paper does.

In the onset year the medical share rises by 2.31 percentage points on a pre-onset mean of 2.63 percent. The leads are flat and about three-fifths of the rise remains the next year. Plague-specific medicine carries the rise. First editions of new works rise by 2.39 percentage points and reprints by 0.23. An onset in a press-free town 20 to 60 kilometers away also raises plague medicine in the printing city.

For titles aimed at Jews and accused witches the onset-year rate ratio is 0.79 and the interval reaches no higher than a rise of 26 percent. Witch trials fall by 0.8 percentage points at onset on a base of 2.6 percent. Persecutions of Jewish communities are too rare to bound tightly. Nothing in the recurrences resembles 1348–1350, when 32 percent of the 471 Jewish communities present in 1347 were attacked or expelled.

Figure 3. The onset estimates against their randomization distributions, as in Figure 12 of the paper. Each city's clean onsets are reassigned at random among its own panel years 500 times. The thick bar spans the middle 95 percent of the reassigned estimates, the thin line the middle 98 percent, and the dot is the actual estimate. Each row has its own scale.

5The prior stock

The response to an outbreak is the shock net of what the city already holds. The paper measures the plague medicine each city printed in the decade before an onset, in years of the city's average output. The coefficient on onset times prior stock is −0.341 with a standard error of 0.135. About a third of a title printed in the prior decade still counted against new demand when the next outbreak arrived. Under geometric decay that implies an annual depreciation rate of 0.21, with a 95 percent interval from 0.09 to 0.56, and an economic life of about five years.

Figure 4. Onset response of plague-specific medicine by the city's prior-decade stock, in titles over the city's average annual output, with the size control held at its mean. The first bin is onsets with no prior stock and the other three are thirds of positive stock. The paper's Figure 14 draws city-clustered intervals. The bands here are the wider of those and Conley intervals computed for this page.

6City explorer

Pick any of the 224 cities to see what its presses printed each year, when plague arrived, and when trials and persecutions took place.

Figure 5. Upper panel, editions printed in the city each year. Middle row, plague outbreaks (rust ticks), clean onsets (rust triangles), witch trials (navy), and persecutions of Jewish communities (gold). Lower panel, medical titles each year, with plague-specific medicine in the darker shade. Hover for the year's counts.

7News as a spot market

The Anabaptist regime at Münster fell to the bishop's army in June 1535. Within the year the story was in print in seven cities. Erfurt and Strasbourg issued a Warhafftiger bericht, Wittenberg a Newe Zeittung, Dresden its own report, and Antwerp and Amsterdam printed the story in Dutch. No press in Münster printed it. The paper's registry holds 384 coercive events with a place and a year, from executions and heresy trials to witch trials, rebellions, and attacks on Jews, together with the cities whose presses reported each one. Pick an event to map who reported it.

Figure 6. The event (star) and the cities whose presses reported it, teal for presses in the event's language and rust for presses in another. Dashed rings mark 250 and 500 kilometers. Distances are great-circle distances. Of the 384 events, 59 were reported only in editions with no known place of printing and draw no reporters.

For each event outside England the paper treats every city that printed news within two years as a candidate reporter, which gives 35,677 city-event pairs. A city within 25 kilometers of an event reported it 20.6 percent of the time, between 25 and 250 kilometers 2.1 percent, and beyond 250 kilometers 0.5 percent. With event and city fixed effects the elasticity of reporting with respect to distance is −0.61, so doubling the distance lowers the probability of a report by about a third. Near a language frontier a same-language town reported an event at 2.4 times the rate of a different-language town at the same distance.

● Same language as the event   ● Another language
Figure 7. Share of candidate cities that reported a coercive event, by distance from the event and by whether the city shared the event's language, on the 35,677 city-event pairs. The vertical scale is logarithmic. Point size is the number of pairs in the bin. From Figure 5 of the paper.

8Doctrine as a backlist

The Malleus maleficarum was first printed at Speyer in 1487. The reprint chains recover 29 editions through 1620 in eight cities, 28 of them with a known place of printing and every one in Latin. Drach at Speyer, Koberger at Nuremberg, and Koelhoff at Cologne printed it six times by 1496. Cologne, Paris, Lyon, and Nuremberg printed it again between 1511 and 1523. Bertano at Venice revived it in four editions between 1574 and 1576. Basse at Frankfurt printed it four times between 1580 and 1600 and Lyon carried it to 1620. A reader in 1600 could buy it from a backlist in Lyon or Frankfurt whatever the timing of plague in his own town.

Figure 8. The 28 placed editions of the Malleus maleficarum printed by the selected year, by city of printing. Circle area is the number of editions and color the year of the city's first edition, dark for the earliest. Arrows follow the order of first editions across cities. From Figure 7 of the paper.

The reprint chains hold 86 doctrine works and 585 coercive-news works. A doctrine work was reprinted 47.7 percent of the time and a news work 28.9 percent. The span from a work's first to its last edition was 9.2 years for doctrine and 0.7 years for news. The odds that a doctrine work appeared in a second language were 22 times the odds for a news work.

Figure 9. Left, the share of works with at least two editions. Right, the years from first to last edition among reprinted works, at each tenth of the distribution, with the median marked large. From Figure 8 of the paper. Reprint links are conservative, so every gap is a lower bound.

9The coverage share

A city's own current output is a share λ = δ / (1 + R) of the content its readers could reach. Here δ is the rate at which a title loses its value each year, which dilutes this year's output against the city's own backlist. R measures the output of other cities within reach, which dilutes it against everyone else's. On one city system common to every product a city's own current output is 1.8 percent of the news its readers could reach and 0.12 percent of the doctrine.

Figure 10. The four products at their estimated depreciation and reach, as in Figure 10 of the paper. Dotted contours hold the coverage share λ constant. The dashed line marks the depreciation of plague medicine estimated in Section 5. Its reach is not calibrated because its demand events are not geocoded. Hover over a product for its numbers.
coverage share λ
of what readers can reach comes from the city's own presses this year
of what readers can reach was printed in other cities
Figure 11. The coverage share for any depreciation and reach. The presets use the estimates of Table 39, Panel C, in the paper.

10Co-movement and its limits

Across decades witch-hunting doctrine and witch executions rise together into the great wave of 1570–1630 and subside together after it. Their correlation is 0.91 in the same decade, 0.80 with print leading by a decade, and 0.65 with print lagging by a decade. Locally the canon does nothing visible. The first arrival of an edition near a town leaves the town's trials flat from 30 years before to 25 years after.

Figure 12. Left, decade totals of witch-hunting doctrine, witch-trial news pamphlets, and witch executions, each indexed to its own maximum. Right, the change in the probability of a witch trial around the first local arrival of the canon, by five-year bin, with city-clustered 95 percent intervals. The paper's last bin pools every year from 30 on and is omitted here. From Figure 9 of the paper.

A printer sizes a run of a durable work to the years it will take to sell. When demand is persistent such a run leads the demand it serves whichever way causation runs. The simulation below sets production equal to the next L years of consumption and the outcome equal to consumption plus noise, so production has no effect on the outcome. Proposition 2 of the paper puts the peak of the correlation at a lead of (L − 1)/2 years. A printer who reissued a doctrine work did so after 4.0 years on average, which puts the predicted lead at 1.5 years.

Figure 13. Left, the simulated correlation between production and the outcome at each lead, averaged over 40 simulations of 4,000 periods with persistence 0.8. The dot marks the peak and the dashed line the predicted peak. The simulation runs in whole years, so its peak falls on a whole year near the prediction. Right, the observed correlation of decade-level doctrine and executions with print leading by k decades. From Figure 11 of the paper.

The aggregate co-movement is what an integrated market produces whether or not reading doctrine raised persecution. In this setting the effect of reading doctrine on persecution is not identified.

11Limits

The event study describes the larger presses. Onsets that enter it sit in presses printing 41.9 titles a year, against 13.9 for those that drop out.

Persecutions of Jewish communities are rare in city-years, and the design cannot rule out an onset-year rise to four times the base rate.

Print aimed at Jews and accused witches mixes durable doctrine with perishable edicts and news. The perishable part is too thin to test on its own, so the bound applies to the category as a whole.

The record of production cannot identify an effect of reading doctrine on persecution, and plague medicine's reach is not calibrated.