New Media, Markets, & Institutional Change Evidence from the Protestant Reformation

New Media, Markets, & Institutional Change
Evidence from the Protestant Reformation
Jeremiah Dittmar
London School of Economics
Skipper Seabold
American University
May 30, 2014
Motivation for Study of Protestant Reformation
Economists & Sociologists
I
Institutions, culture, and religion – are deep, persistent
determinants of behavior and well-being.
Historians
I
Dynamics of institutional change are very important.
I
Reformation is one of the most profound social changes.
I
New media are key to breach of Catholic monopoly.
I
Big data challenge: Hard to measure ideas in media.
History “rhymes”
I
New media, Arab Spring, Occupy Wall Street...
Today: Evidence from the Reformation
1. Measure of Protestant content in media.
I
I
I
Firm-level panel: all known books & pamphlets 1454-1600.
110,000+ varieties by 1,100+ firms in 146 German cities.
Classify content with estimators for high-dimensional data.
2. Competition and institutions as determinants of diffusion.
I
I
I
Cities with more firms ⇒ more Protestant media.
Competition matters more where political freedom is low.
Timing of printer deaths ⇒ shocks to city competition.
3. Institutional change.
I
I
1,000+ municipal Reformation laws.
Laws design & set-up first mass public education system.
5
Big Picture: Index of Religious Content Across Time
Religious Books in Latin
1
.75
.75
Religion Index
Religion Index
Religious Books in German
1
.5
.25
1600
0
1475
.5
.25
1500
1525
1550
1575
1600
0
1475
1500
Religion Index: 1 = Protestant. 0 = Catholic.
1525
1550
1575
1600
Historians on the Role of Print Media and Printers
“No printing, no Reformation.”
– Moeller, Stadt und Buch (1979)
“The decision to print or not to print a particular book or tract
could have an immediate effect on political and religious events
and, in a time of rapid change, on institutions.
The most striking example of [printers’] influence can be seen in
the religious publication of the pivotal years of the Reformation.
No one could foretell Luther’s eventual success. The
magistrates had by no means adopted a favorable position
toward the Reform.”
– Chrisman, Lay Culture, Learned Culture (1981)
Fact 1: Transmission Via Printed & Spoken Word
I
Prices: typical 32 page pamphlet = 1/3 daily wage
I
Literacy: 30% in cities, 5% overall in early 1500s (?)
I
Transmission:
I
Media impacts “opinion leaders”.
I
Opinion leaders transmit ideas orally to broad public.
“Why does the pope, whose wealth today is greater than the
wealth of the richest Crassus, build the basilica of Saint Peter
with the money of poor believers rather than with his own
money?”
– Martin Luther (86th thesis, 1517)
Fact 2: Ideas & Institutions Diffuse First at City Level
Figure: Cities of German-Speaking Europe circa 1500
Fact 3: Reformation Emerges as Popular Movement
“It is undeniable that the Wittenberg movement was borne on a
wave of popular enthusiasm. It outran the city magistrates’
ability to control it, and finally forced them to act even against
the will of the Elector [the regional prince], who had prohibited
any innovations in church matters.”
– Scribner (1979)
“As a rule neither the city patricians nor the local princes
showed any sympathy for the Reformation in the crucial period
in the 1520s and early 1530s; they identified themselves with
the old Church hierarchy...Popular agitation on a broad social
base led to the formation of a ‘burgher committee’.”
– Cameron (1991)
Fact 4: Free Entry and Minimal Regulation
“The industry was free to develop without regulation by
governments, princely houses or the Church, nor is there any
evidence that any restrictions were imposed by guilds.” (Füssel
2005)
“Trades that became large after the list of officially approved
guilds was drawn...escaped guild regulation...Printing is the
most obvious example.” (Nicholas 2003)
“The new occupations tied to printing fell outside the framework
of the old guilds.” (Barbier 2006)
Data
Raw data – all known books & pamphlets 1454-1600
I
Universal Short Title Catalogue. 110,000+ publications.
I
No direct meta-data on printers/firms.
I
37 classified subjects.
I
“Religion” counts for 35% – has no sub-classification.
Research
I
Code up the firm that produced each variety.
I
Identify religion for 455 authors who wrote 18,000 books.
Unit of analysis
I
The book/pamphlet title as a “variety”.
I
Variety i by firm j in city k in year t.
How Firm Level Data are Constructed
From Inscriptions on Individual Books
Printer Inscription
on Front Page of Book or Pamphlet
Johann Schönsperger
Johann Schönsperger & Thomas Rüger
Gedruckt und volendt von Anna Rügerin
in der keyserlichen stat Augspurg
Heinrich Steiner von Augsburg
Heinrich von Augsburg Steiner
Heinrich von Augsburg Steiner & haer.
Erhard Oeglin
excudebat Heinrich von Augsburg Steiner
apud Heinrich von Augsburg Steiner
Standardized
Firm Name #1
Johann Schönsperger
Johann Schönsperger
Thomas Rüger
Heinrich Steiner
Heinrich Steiner
Heinrich Steiner
Widow or
Heir #1
Standardized
Firm Name #2
Widow or
Heir #2
Thomas Rüger
Yes
Erhard Oeglin
Heinrich Steiner
Heinrich Steiner
This example: 8 books printed in Augsburg ⇒ 4 firms
Complete data: 110,000+ varieties ⇒ 1,000+ firms
99% of historic books and pamphlets identify printer
Yes
Print Media in German-Speaking Europe
Decade
Starting
1450
1460
1470
1480
1490
1500
1510
1520
1530
1540
1550
1560
1570
1580
1590
Cities
Printing
Firms
Printing
1
4
23
39
36
43
40
58
53
55
67
68
70
79
85
2
14
321
599
721
673
767
1,045
1,099
1,116
1,383
1,622
1,606
1,741
2,187
Media Output: Book and Pamphlet Varieties
Total
Religious Protestant Catholic
Varieties
Varieties
Author
Author
3
42
1,645
3,136
4,206
3,989
5,389
11,171
7,854
8,187
9,412
10,669
9,963
13,172
16,761
3
28
867
1,565
1,531
1,443
1,844
7,326
3,344
3,677
4,448
4,830
3,829
5,012
6,141
0
0
1
4
4
14
399
3,825
1,600
1,924
2,291
2,131
1,236
1,152
795
0
0
6
0
2
80
190
522
448
339
391
388
269
183
103
Examples of Text Data to Model and Classify
Protestant title — by Martin Luther, printed in Wesel
The last wil and last Confeßion of martyn Luthers faith
concerning the principal articles of religion which are
in controversy, which he wil defend & mainteine until
his death, agaynst the pope and the gates of hell.
Catholic title — by John Old, printed in Emden
A Confeßion of the most auncient and true christen
catholike olde belefe accordyng to the ordre of the XII
Articles of our common crede, set furthe in Englishe to
the glory of almightye God, and to the confirmacion of
Christes people in Christes catholike olde faith.
Religion Index of Media Content
Idea:
I
Two types of author – Protestant and Catholic.
I
Distribution of language changes with religion.
Hand code the religion of 400+ known authors (18,000+ titles).
Use publications by authors with known religion to identify
divergent “religious phrases” (5,000+ phrases).
1st stage: Measure Protestantism in religious phrases using
statistical model for sparse multinomial data.
2nd stage: predict religious content where ideology unknown.
Model: Taddy’s (2013) multinomial inverse regression.
Religion Index
Big Picture: Get weights that relate phrases to religion.
Then predict ideology of religious books with unknown authors.
Identify a vocabulary (W ) of religious phrases employed
authors with known religion.
Imagine vocabulary is: [‘Luther’, ‘Holy Pope’, ‘Dr. of Theology’,
‘catholic old belief’, ‘gates of hell’]
I
Lutheran text might be: [85, 0, 92, 0, 57]
I
Catholic text might be: [73, 90, 8, 82, 0]
To construct vocabulary
I
Use log-odds measure of Monroe et al. (2008).
I
Less Type I & II error than χ2 of Gentzkow & Shapiro
(2010)
I
Cardinality of support: |W | = 5000+ phrases
Divergent Language Used to Construct Religion Index
15
Log-Odds Ratio - Informative Dirichlet Prior - German
sermon
christi
unde
eine
am
herrn
fuer
christen
de
10
van
auslegung
dat
sol
lutheri
predigten
epistel
psalm
jhesu
unterricht
sermon
unde
5
ln(Õ(wP) /Õ(wC) )
christi
eine
fuer amherrn
christen
van
auslegung
dat
solde
lutheri
predigten
epistel
psalm
jhesu
unterricht
jungen
jungen
0
wirdt
hohen
heiligen
bern
gelesen
5
bern
secten
catholisch
rechter
altars
predigen
bißalten
warheit
catholischer
catholische
predigen
biß
kirchen
oder
heiligen
kirchen
oder
secten
alten
catholisch
rechter
postill
evangelischer
catholischen
10
15
hohen wirdt
gelesen
altars
warheit
catholischer
catholische
postill
evangelischer
catholischen
100
101
102
Frequency of words
103
104
Religion Index
1. Estimate sufficient reduction score – scalar summary
statistic that preserves information on religion
I
I
I
I
Document X is multinomial with parameter vector q
Estimate factor loadings ϕ relating phrase counts to q
Use loadings to estimate “sufficient reduction score”:
zi = ϕ0 f i where f i = x i /mi
“Sufficiency”: religion yi independent of phrase counts, xi
and document lengths mi , given zi .
2. Forward regression – relationship between religion and
reduction score for authors with known religion
P(religion = protestant) = yi = [exp(α + βzi )]−1
3. Second stage prediction – predict religion for unknowns
I
Use α̂, β̂, and SR zi predicted out of sample.
Note: We have also estimated with controls for time and region FE
that shift the distribution of language.
Does the Classifier Work Out of Sample?
1.0
Martin Luther
Pr(Protestant==1)
0.8
0.6
0.4
0.2
0.0
Johannes Eck
Protestant
Catholic
Use 80% of the knowns as “training” corpus.
Hold out remaining 20% plus Martin Luther & Johannes Eck.
Predict on held outs. 86% success rate in random subsamples.
5
Question: What determines local variations in diffusion
behind this big picture aggregate story?
Religious Books in Latin
1
.75
.75
Religion Index
Religion Index
Religious Books in German
1
.5
.25
1600
0
1475
.5
.25
1500
1525
1550
1575
1600
0
1475
1500
Religion Index: 1 = Protestant. 0 = Catholic.
1525
1550
1575
1600
Determinants of Diffusion
Historians emphasize
I
Institutional status of city: Free Imperial vs. subject to lord.
I
“Development” of print industry, Germanic humanism.
I
Preferences of territorial princes/lords.
I
Censorship was weak/endogenous: “as much a product of
public opinion as a force acting upon it” (Creasman 2012)
Market competition in this research
I
Number of firms competing in city before Reform.
I
Trade costs ⇒ salience of city as unit (Dittmar 2013).
Institutions in this research
I
Territory level propensity.
I
Within-territory variation in institutional autonomy of cities.
Firms and City Populations on Eve of Reformation
64
Firms Active 1508-1517
32
Strasbourg
Augsburg
Koeln
Basel
16
Leipzig
8
Wien
Mainz
Erfurt
4
Wittenberg
Tuebingen
München
Münster
2
1
Trier
1
2
4
8
16
City Population in 1500 (Thousands)
32
64
Cities with population unknown omitted here. Several cities have identical
numbers of firms and population. Marker scale is titles per firm.
Pre-Reformation Institutions
Principality level institutions
I
30+ principalities.
I
Electoral Palatinate, Duchy of Württemberg, Swiss
Confederation, Duchy of of Saxony, ...
Within principality borders *
I
Cities subject to feudal “lord rule” vs. “free cities.”
I
Electoral Palatinate: Landau was a free city.
Kaiserslautern, Heidelberg, Mannheim, Oppenheim, and
Zweibrücken were subject to the lord Elector.
I
The Swiss Confederation: Basel, Schaffhausen, and Sankt
Gallen were free. Bern, Solothurn, and Zürich were not.
Empirical Strategy
1. Model exposure to Protestant media with hurdle model
I
I
Counts: “excess” zeros & over-dispersion relative to
Poisson ⇒ Akaike criteria ⇒ model w/zero-truncated NB.
Two processes (i) any Protestant (ii) count of Protestant.
2. Exogenous variation in competition – manager deaths
3. Placebo – competition matters for Protestant, not Catholic
4. High frequency data
I
I
Evolution of relation b/w competition and Protestant media.
Comparison to literature using distance-based IV.
Determinants of “Any Protestant” – Marginal Effects
Dependent Variable is Indicator for Any Protestant Media
Any
Any
Any
Any
Variable
Protestant Protestant Protestant Protestant
(1)
(2)
(3)
(4)
(5)
Firms 1508-1517
0.13***
0.17***
0.15***
0.01
(0.04)
(0.06)
(0.05)
(0.04)
Indicator: Lord Rule
0.10
-0.04
-0.04
0.08
(0.08)
(0.09)
(0.08)
(0.08)
Firms 1508-1517 x Lord Rule
0.22***
(0.05)
Indicator: Any Firms Pre-1517
0.05
(0.15)
Distance to Wittenberg
0.07**
-0.00
-0.00
0.07**
(0.03)
(0.04)
(0.04)
(0.03)
Latin Media pre-1517
0.24
0.07
0.06
0.32
(0.19)
(0.33)
(0.31)
(0.22)
Vernacular Media pre-1517
0.02
0.11
0.11
0.01
(0.11)
(0.19)
(0.19)
(0.10)
Indicator: Hanseatic
-0.03
0.16
0.16
-0.02
(0.10)
(0.12)
(0.12)
(0.10)
Principality Fixed Effects
Yes
Yes
Observations
268
234
234
268
Any
Protestant
(6)
0.10*
(0.05)
-0.05
(0.09)
0.19
(0.12)
-0.00
(0.04)
0.16
(0.34)
0.08
(0.19)
0.15
(0.12)
Yes
234
Any
Protestant
(7)
0.09
(0.06)
-0.06
(0.08)
0.18*
(0.10)
0.05
(0.15)
-0.00
(0.04)
0.15
(0.32)
0.08
(0.18)
0.15
(0.12)
Yes
234
Logit marginal effects. Standard errors clustered by principality. Controls for
population in bins (unknown, 1k, 2-5k, 6-10k, 11-25k, 25k+).
Determinants of “Count Protestant”
Dependent Variable: Protestant Varieties Post-1517
Protestant Protestant Protestant Protestant
Variable
Varieties
Varieties
Varieties
Varieties
(1)
(2)
(3)
(4)
(5)
Firms 1508-1517
0.05**
0.05**
0.04**
0.05*
(0.02)
(0.02)
(0.01)
(0.03)
Indicator: Lord Rule
0.10
0.16
0.46
0.02
(0.26)
(0.62)
(0.62)
(0.24)
Firms 1508-1517 x Lord Rule
0.07*
(0.04)
Indicator: Any Firms Pre-1517
1.07**
(0.45)
Distance to Wittenberg
-0.08
-0.12
-0.07
-0.08
(0.05)
(0.18)
(0.18)
(0.05)
University 1517
0.88***
1.05***
0.69*
0.74***
(0.20)
(0.37)
(0.36)
(0.22)
Latin Media pre-1517
-0.02
-0.03
-0.05
-0.03*
(0.02)
(0.03)
(0.04)
(0.02)
Vernacular Media pre-1517
0.05
-0.00
0.05
0.04
(0.07)
(0.09)
(0.05)
(0.08)
Indicator: Hanseatic
0.01
-0.12
-0.10
-0.00
(0.21)
(0.24)
(0.20)
(0.21)
Principality Fixed Effects
Yes
Yes
Observations
116
116
116
116
Protestant
Varieties
(6)
0.05*
(0.03)
0.08
(0.55)
0.08
(0.06)
-0.12
(0.18)
0.94***
(0.34)
-0.05
(0.04)
-0.01
(0.10)
-0.11
(0.27)
Yes
116
Protestant
Varieties
(7)
0.04**
(0.02)
0.47
(0.63)
-0.00
(0.04)
1.07**
(0.47)
-0.07
(0.18)
0.70**
(0.35)
-0.05
(0.04)
0.05
(0.05)
-0.10
(0.20)
Yes
116
Model counts w/zero-truncated neg binomial. SE’s clustered by principality.
Manager Deaths as Shocks to Competition
Use variation induced by timing of deaths
I
Master printer has hard-to-replace skills & contacts.
I
Anti-competitive arrangements widespread but fragile.
I
Median city has 3.5 firms in any year.
I
Deaths matter for within city competition.
I
City salient unit because transport costs limit trade.
“It is difficult to overestimate the disruption caused by the death
of a master printer.” – Parker (1996)
Use evidence on network links of dying capitalists
“in 1535 he entered a printing association with his brother in law
Robert Winter, Balthasar Lasius, and Thomas Platter the Elder.”
The Distribution of Manager Age at Death
.04
Active Printers
Inactive Printers
Density
.03
.02
.01
0
30
40
50
60
Age at Death
70
80
90
New Entrants in Annual City-Level Data
The Impact of Manager Deaths
Variable
(1)
Manager Death
Lagged Firms
Year Fixed Effects
City Fixed Effects
City-Decade Fixed Effects
Observations
Probit Marginal Effects
Probability
Probability
of an Entrant
of an Entrant
(2)
(3)
0.12***
0.13**
(0.05)
(0.05)
-0.01*
-0.01*
(0.00)
(0.00)
Yes
Yes
Yes
4343
4343
Identification idea: “Not Augsburg in 1510s, but the precise year in Augsburg
in the 1510s when a manager dies.” SE’s clustered by city or city-decade.
O
Num
Fi
(
0.3
(0.
0.9
(0.
Y
Y
46
Number of Firms in Annual City-Level Data
The Impact of Manager Deaths
Variable
(1)
Manager Death
Number of
Firms
(2)
0.26***
(0.07)
Manager Death x Publications
Number of
Firms
(3)
0.26**
(0.13)
0.00
(0.00)
Manager Death x Business Ties
Lagged Firms
Year Fixed Effects
City Fixed Effects
Observations
0.92***
(0.02)
Yes
Yes
4408
0.92***
(0.02)
Yes
Yes
4408
“Publications” is # varieties published over 4 years before death.
“Business Ties” is # of inter-firm links. SE’s clustered by city.
Number of
Firms
(4)
0.17*
(0.09)
0.41***
(0.08)
0.91***
(0.02)
Yes
Yes
4408
The Current Impact of Lagged Manager Deaths
The Outcome is Current Number of Firms in a City
1.5
Current Impact on City-Level Firms
1.25
1
.75
.5
.25
0
-.25
-.5
-3
-2
-1
0
1
2
3
Year Relative to Firm-Level Shock
4
5
Controls for city and year FE, and for longer lag of # firms than previous slide.
Manager Deaths and Change in Number of Competing
Firms in a City in the 10 Years Before Luther
Change in # Firms 1498-1507 to 1508-1517
5
0
-5
-10
0
1
Printer Deaths 1508-1517
2
Markers scaled to reflect number of firms in prior period (i.e. 1498-1507).
Manager Deaths and Number of Competing Firms
Deaths and Firms in the 10 Years Before Luther
Dependent Variable: Number of Firms Active 1508-1517
All
All
All
All
Printing
Variable
Cities
Cities
Cities
Cities
1498-1507
(1)
(2)
(3)
(4)
(5)
(6)
Printer Deaths 1508-1517
1.04**
1.20**
1.30***
1.40***
1.08**
(0.44)
(0.54)
(0.31)
(0.43)
(0.50)
Firms 1498-1597
-0.36**
-0.35**
-0.23
(0.16)
(0.17)
(0.20)
Distance to Wittenberg
0.07
0.13
0.07
0.13
0.60
(0.05)
(0.10)
(0.05)
(0.08)
(0.37)
University 1517
-0.39
-0.54
0.28
0.12
1.14
(0.77)
(0.91)
(0.80)
(0.94)
(1.45)
Latin Media pre-1517
-0.15
-0.12
0.11
0.12
0.03
(0.18)
(0.21)
(0.33)
(0.35)
(0.34)
Vernacular Media pre-1517
0.24
0.10
0.86***
0.73***
0.88***
(0.40)
(0.44)
(0.22)
(0.25)
(0.26)
Indicator: Lord Rule
0.14
-0.02
0.02
-0.05
0.48
(0.10)
(0.12)
(0.09)
(0.11)
(0.54)
Indicator: Hanseatic
0.11
0.10
0.10
0.11
1.06
(0.08)
(0.12)
(0.09)
(0.12)
(1.18)
Principality Fixed Effects
Yes
Yes
Observations
286
286
286
286
43
Standard errors clustered by principality.
Printing
1498-1507
(7)
1.88
(1.25)
-0.70**
(0.27)
-0.35
(0.96)
-0.62
(1.80)
0.88***
(0.29)
0.93
(0.66)
-0.64
(2.31)
-1.61
(1.52)
Yes
43
Determinants of “Any Protestant” – Marginal Effects
IV Probit with Firms Induced by Manager Deaths
Dependent Variable is Indicator for Any Protestant Media
Data on All Cities
Cities Printing 1498-1507
Variable
IV Firms
IV Increase IV Increase
IV Firms
IV Increase
(1)
(2)
(3)
(4)
(5)
(6)
Firms 1508-1517
1.42***
0.69***
(0.18)
(0.06)
Increase in Firms 1508-1517
1.68***
1.92***
0.77***
(0.19)
(0.25)
(0.06)
Firms 1498-1597
1.06***
1.26***
0.46***
(0.21)
(0.21)
(0.10)
Distance to Wittenberg
0.09
0.06
-0.06
0.03
-0.01
(0.07)
(0.06)
(0.07)
(0.21)
(0.23)
Latin Media pre-1517
-1.59**
-1.86**
-3.06***
-0.91**
-0.88**
(0.74)
(0.94)
(0.93)
(0.40)
(0.43)
Vernacular Media pre-1517
-0.84
0.95
3.43***
-0.20
0.63
(1.21)
(1.69)
(1.07)
(0.48)
(0.69)
Indicator: Lord Rule
0.08
-0.06
-0.13
-0.07
-0.32
(0.14)
(0.17)
(0.18)
(0.38)
(0.46)
Indicator: Hanseatic
-0.08
-0.11
0.27
-0.19
-0.13
(0.19)
(0.15)
(0.22)
(0.59)
(0.63)
Principality Fixed Effects
Yes
Observations
268
268
234
41
41
Binary “any protestant.” Probit marginal effects. Standard errors clustered by
principality. Baseline was β̂ = 0.17. IV is deaths of printer in city 1508-1517.
Determinants of “Count Protestant”
GMM with Lagged Firms & Manager Deaths as IV
Variable
(1)
Firms 1508-1517
Indicator: Lord Rule
Distance to Wittenberg
University 1517
Latin Media pre-1517
Vernacular Media pre-1517
Indicator: Hanseatic
Observations
Dependent Variable: Protestant Varieties Post-1517
IV: Lagged Firms
IV: Lagged Firms & Deaths
All Cities
Print Cities
All Cities
Print Cities
(2)
(3)
(4)
(5)
0.09**
0.08**
0.08***
0.08**
(0.04)
(0.04)
(0.03)
(0.03)
0.13
0.16
-0.04
0.11
(0.25)
(0.30)
(0.17)
(0.23)
-0.09*
-0.04
-0.09*
-0.05
(0.05)
(0.10)
(0.05)
(0.09)
0.75***
0.73***
0.88***
0.76***
(0.22)
(0.25)
(0.19)
(0.21)
-0.04
-0.02
-0.03
-0.02
(0.03)
(0.02)
(0.02)
(0.02)
-0.07
0.10
-0.05
0.11
(0.09)
(0.07)
(0.08)
(0.07)
-0.02
0.60***
0.07
0.60***
(0.20)
(0.17)
(0.21)
(0.18)
116
33
116
33
IV: Deaths
All Cities
Print Cities
(6)
(7)
0.30**
0.08**
(0.12)
(0.03)
0.16
0.11
(0.25)
(0.23)
-0.10*
-0.05
(0.05)
(0.09)
0.37
0.76***
(0.40)
(0.21)
-0.08*
-0.02
(0.05)
(0.02)
-0.54*
0.11
(0.29)
(0.07)
-0.16
0.60***
(0.25)
(0.18)
116
33
Outcome is count Protestant. GMM estimates. IV is deaths of printer in city
1508-1517. Standard errors clustered by principality. Baseline β̂ = 0.05.
Over-identification test for (4) has p = 0.44, in (5) p = 0.63 [Hansen’s J]
Natural Questions: What about the relationship
between competition and Catholic media? Geographic
diffusion? What happens over time?
Competition measured by number of firms
I
Religious media in general & Protestant differential.
Distance to Wittenberg
I
Religious media in general & Protestant differential.
Interaction of competition and distance (firms x distance)
I
Religious media in general & Protestant differential.
Observe how all these relationships vary year-by-year.
Restrict to cities printing on eve of Reform (1508-1517).
Estimation Idea: Use interaction terms to document
evolution of relationship between religious media,
competition, and distance at city level
Yikt =
1600
X
t=1500
+
1600
X
t=1500
1600
X
t=1500
(βtf Dt · firmsi )
|
{z
}
+
t=1500
religious media on competition
(βtd Dt · disti )
|
{z
}
religious media on distance
(β fd Dt · firmsi · disti )
| t
{z
}
1600
X
+
t=1500
religious media on firms x distance
+
1600
X
(βtfp Dt · Protk · firmsi )
|
{z
}
Protestant slope on competition
(βtdp Dt · Protk · disti ) +
|
{z
}
Protestant slope on distance
1600
X
t=1500
(βtfdp Dt · Protk · firmsi · disti )
{z
}
|
Protestant slope on firms x distance
+ δt + θi + it
Estimate via OLS and negative binomial regression.
Plot annual slopes β̂tf , β̂tfp , ...
Firms, Distance-to-Luther, & Religious Media
Baseline & Protestant Differential – Negative Binomial
Firms
All Religious Media
Distance
All Religious Media
0.40
4.00
0.20
2.00
0.00
0.00
Firms x Distance
All Religious Media
0.20
0.10
0.00
-0.20
-2.00
-0.40
-4.00
-0.10
1500 1525 1550 1575 1600
1500 1525 1550 1575 1600
1500 1525 1550 1575 1600
Firms
Protestant Media
Distance
Protestant Media
Firms x Distance
Protestant Media
0.60
2.00
0.40
1.00
0.20
0.00
0.00
-1.00
-0.20
-2.00
0.05
0.00
-0.05
-0.10
1500 1525 1550 1575 1600
-0.15
-0.20
1500 1525 1550 1575 1600
1500 1525 1550 1575 1600
Religious Media and Legal Reform
1
Religion Index for Media
.8
.6
.4
.2
0
1475
1500
1525
Cities that Pass Laws
1550
1575
1600
Cities that do not Pass Laws
The Timing of City Level Legal Reforms
Laws Passed Per Year
8
6
4
2
0
1500
1520
1540
1560
1580
1600
1620
Cities Passing First Reformation Law
Total Reformation Laws (5 Year Moving Average)
1640
Legal Reform at the City Level
I
Laws written/passed by city magistrates.
I
Set up first mass experiment in public education.
I
Annual audit & assessment of schools/teachers.
Our current data codes: “when laws passed”.
We are currently coding the provisions:
I
“both boys’ & girls’ schools”, “university scholarships”,
“poor law”, “common chest for welfare expenses”, etc.
Key Source: Sehling Evangelischen Kirchenordnungen
(21 volumes 1902-1910). Other sources for outside Germany.
Legal Reform Across Cities
:
# of Reformation Ordinances by City
Bairoch City
0
20
40
80
120
Figure: Red
cities pass
Reformation Law.
Blue cities do not.
Sehling
Reformation
Ordinances
Bairoch Cities w/ Ordinances
Bairoch Cities w/o Ordinances
160
Miles
Robinson Projection
Central Meridian: 30.00
Pre-1517 Determinants of City-Level Legal Reform
Dependent Variable is Indicator for Cities that Instituted Legal Reform
P(Reform) P(Reform) P(Reform) P(Reform) P(Reform)
-0.00
-0.00
-0.00
-0.00
0.00
(0.01)
(0.01)
(0.01)
(0.01)
(0.01)
Firms 1510-1517 x Lord Rule
0.03*
0.04**
(0.02)
(0.01)
Indicator: Printing 1510-1517
Variable
Firms 1510-1517
Distance to Wittenberg
University in 1517
Latin Books pre-1517
Vernacular Books pre-1517
-0.12***
(0.02)
-0.31***
(0.09)
-0.00
(0.01)
0.05
(0.03)
-0.13***
(0.02)
-0.22**
(0.10)
-0.00
(0.01)
0.07**
(0.03)
-0.17*
(0.10)
142
142
Indicator: Lord Rule
Principality Fixed Effects
Observations
-0.08
(0.05)
-0.14
(0.11)
-0.02
(0.01)
0.12***
(0.02)
0.01
(0.16)
Yes
142
-0.12***
(0.02)
-0.29**
(0.11)
-0.01
(0.01)
0.07**
(0.03)
-0.19*
(0.09)
142
-0.07
(0.05)
-0.22*
(0.12)
-0.02**
(0.01)
0.12***
(0.02)
-0.01
(0.15)
Yes
142
P(Reform)
0.01
(0.01)
0.05***
(0.02)
-0.07
(0.10)
-0.07
(0.05)
-0.20
(0.12)
-0.02***
(0.01)
0.12***
(0.02)
-0.01
(0.16)
Yes
142
Figure: German-speaking cities with population observed in 1500.
Controls for population. Standard errors clustered by principality.
Media Exposure and Legal Change
Legal Reform: simple survival model ⇒ illustrative correlations.
Probability: city i passes first Reformation law at time t,
conditional on not yet having legal reform.
Treatment: cumulative exposure to Protestant/Catholic media.
Time dependence: polynomials of time
(similar with cubic splines, lowess smoothed)
Generic challenge: limited dependent variable models
inefficient and inconsistent if there’s unobserved spatial
correlation in the error.
Media Exposure and Legal Change
log2 (Cumulative Protestant)t−1
log2 (Cumulative Catholic)t−1
log2 (Cumulative Vernacular)t−1
Distance from Wittenberg
Imperial Free City
1.27**
(0.15)
0.60***
(0.07)
1.35**
(0.20)
0.99***
(0.00)
2.42*
(1.09)
W 0 Yt−1 (Neighbors’ Law)
N
6770
1.26*
(0.15)
0.59***
(0.07)
1.37**
(0.20)
0.99***
(0.00)
2.41**
(1.06)
3.35**
(1.63)
6770
“Double Protestant ⇒ 23% higher chance of Reform.”
“Double Catholic ⇒ 50% lower chance of Reform.”
Conclusions & Future Directions
New measure of media content
Preliminary evidence: media market competition, diffusion of
ideas that lead to institutional change.
To come:
I
Exploit firm shocks (manager deaths) in time series.
I
Cross border media spillovers and legal reform.
I
Impact of legal & educational institutions of Reformation.
Comments/criticisms kindly solicited!
[email protected]