Lecture 1
Introduction
The Malthusian Era
Matti Sarvimäki
Economic History
23 February 2015
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Why history?
Motivation 1: Consumption value
intellectual curiosity
entertainment
aspiration to appear “civilized”
Matti Sarvimäki
Economic History
1: Introduction
1 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Why history?
Motivation 1: Consumption value
intellectual curiosity
entertainment
aspiration to appear “civilized”
This is not why we are here
though all means to get motivated are allowed
Matti Sarvimäki
Economic History
1: Introduction
1 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Why history?
“[Ancient Greeks] saw the future as something that came upon
them from behind their backs with the past receding away before
their eyes. When you think about it, that’s a more accurate
metaphor than our present one. Who really can face the future? All
you can do is project from the past, even when the past shows that
such projections are often wrong” (Robert M. Pirsig: Zen and the Art of
Motorcycle Maintenance, 1974, afterword)
Matti Sarvimäki
Economic History
1: Introduction
2 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Why history?
Motivation 2: Improving decision making
avoiding past mistakes
improving prediction
testing formal models
estimating important parameters
We are here to make you better economists
most of the material published in top economics journals
this focus has both limitatations and strengths; you should not
think this as a substitute for “traditional” econ. history courses
Matti Sarvimäki
Economic History
1: Introduction
3 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
The Question: How did we become so well-off?
An anecdote illustrating how much better
off we are than people just 150 years ago:
Prince Albert, the husband of Queen
Victoria, died at the age of 42 in 1861. The
likely cause of death was typhoid. It is a
bacterial disease typically transmitted by
drinking water being polluted by sewage. It
was common in the 19th century cities, but
has virtually disappeard from rich countries
due to vaccinations and better public
sanitation such as chlorination of drinking
water and the building of effective sewerage
systems.
Matti Sarvimäki
Economic History
1: Introduction
4 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
What is to be explained?
Mortality under age one per 1,000 children in Finland (Statistics Finland)
400 350 300 250 200 150 100 50 0 1750 Matti Sarvimäki
1800 1850 Economic History
1900 1950 1: Introduction
2000 5 / 37
What is to be explained?
Real wages of English building workers (Clark 2005, JPE)
1308
journal of political economy
Fig. 1.—Builders’ real day wages, 1209–2004 (source: table A2)
Based on 46,000 quotes of day wages, 90,000 quotes of the prices of 49 commodities,
and 20,000 quotes of housing rents
wage observations and 110,000 observations on prices and housing
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
What is to be explained?
GDP per Capita 1–2008 (www.ggdc.net/maddison)
GDP per capita (thousands of 1990$)
30
25
20
15
10
5
0
0
200
400
Western Europe
Matti Sarvimäki
600
800
1000
1200
Western Offshotts
Economic History
1400
1600
Asia
1: Introduction
1800
2000
Africa
7 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
What is to be explained?
GDP per capita in Finland, Sweden and the U.S. (www.ggdc.net/maddison)
GDP per capita (thousands of 1990$)
32
1917
1939
1970
16
Hungary (2008)
8
4
Morocco (2008)
2
Ghana (2008)
1
1820
1840
1860
1880
1900
Finland
Matti Sarvimäki
Economic History
1920
Sweden
1940
1960
1980
2000
United States
1: Introduction
8 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Explaining Economic Growth
Macroeconomics I: Growth and Cycles
Aggregate production is
Y = Af (K , L, X )
where A is total factor productivity and f (·) is a constant returns
to scale production function (factors: capital, labor, land).
Standard growth models provide a valuable framework to think
systematically about the proximate causes of growth
Matti Sarvimäki
Economic History
1: Introduction
9 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Explaining Economic Growth
This course
If technology, physical capital and human capital are so
important, why do they differ across locations and time?
In this course, we focus on the fundamental causes of
growth
i.e. potential reasons for why the proximate causes vary
complements Macro 1
Matti Sarvimäki
Economic History
1: Introduction
10 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
We will use four complementary approaches
1
Narrative history
the “story” of what happened and why
2
Formal models
internally consistent simplifications of the complex reality
3
Descriptive empirical work
careful measurement of what actually happened
4
Causal empirical work
(natural) experiments and/or quantitative economic models
used to compare counterfactual states of the world
Matti Sarvimäki
Economic History
1: Introduction
11 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Outline of the course
1
The Malthusian Era
2
Fundamental causes of growth
1
2
3
4
3
Luck
Geography
Culture
Institutions
Innovation and crises
1
2
4
(today)
Technology
Finance
Unleashing talent
1
2
3
Matti Sarvimäki
Migration
Social mobility
Marriage, family and work
Economic History
1: Introduction
12 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Change in schedule
Next lecture rescheduled to FRIDAY 27th
Matti Sarvimäki
Economic History
1: Introduction
13 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Grading and Requirements
Essays (50%)
Two essays about paper-pairs listed in the syllabus
Max. 1,000 words, need to answer:
What are the “take-aways”?
What are the key assumptions? Are they plausible (and why)?
How do the papers contradict/complement each other?
The essays are due to March 12th and April 10th by email to
[email protected]
Final exam (50%)
You are expected to know the material covered in the lectures
Passing the exam required for passing the course
Lecture notes at: www.aalto-econ.fi/sarvimaki/courses.html
April 7th, 9am (retake: May 22nd, 2pm)
Matti Sarvimäki
Economic History
1: Introduction
14 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
The Malthusian model
In order to understand why some
countries started to grow rapidly
about two centuries ago, we first
need to understand why they failed
to do so earlier
The main explanation is due to
Thomas Malthus, whose book An
Essay on the Principle of
Population was first published in
1798.
Reverend Thomas R. Malthus, 1766–1834.
One of the many English clergymen of his
time, who are still remembered today.
Matti Sarvimäki
Economic History
1: Introduction
15 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
The Malthusian model
Malthus’ model builds on the key assumptions that:
population increases with per capita income
fixed supply of land leads to negative returns to scale
technological progress is slow (implicit)
Matti Sarvimäki
Economic History
1: Introduction
16 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
The Malthusian model
Malthus’ model builds on the key assumptions that:
population increases with per capita income
fixed supply of land leads to negative returns to scale
technological progress is slow (implicit)
Given these assumptions, the only way to achieve a permanent
increase in per capita income is to restrict population
e.g. technological innovation increases productivity → wages
increase → population grows → land/labor ratio decreases →
wages decrease back to subsistance level
Matti Sarvimäki
Economic History
1: Introduction
16 / 37
Steady-State in the Standard Malthusian Model
Voigtländer and Voth (2013,
776 RESTUD)
REVIEW OF ECONOMIC STUDIES
Steady states with "Ho
birth rate
C
Death/Birth rate
Death/Birth rate
Steady state in the standard Malthusian model
E
0
EU
death rate
wC
Wage
w0
Figure 1
“Death rates are downward sloping Steady
in income,
and
birthMalthusian
rates are
either
flat“Horsemen
or
states in the
standard
model
and with
effec
upward sloping. This generates a unique steady state (C) that pins down wages
and population size. Decreasing marginal returns to labour set in quickly as
it converges
EH . isThis
equivalentfactor
to a “ratchet
effect”: Wage
population grows because
fixedtoland
an is
important
of production.
A gains aft
6
became
permanent.
decline in population
can raise
wages, moving the economy to the right of C.
crucial feature to obtain multiple steady states in Figure 1 is an u
However, the increase The
in output
per capita is only temporary. Birth rates now
of the death schedule. This reflects the historical realities of early modern E
exceed death rates and population grows, which in turn will depress wages—the
to Malthus (1826), factors reducing population pressure include “vicious cu
“Iron Law of Wages”to holds.”
women, great cities, unwholesome manufactures, luxury, pestilence, and
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
The Malthusian model: Discussion
Malthus presented the first coherent theory about how limited
resources constraint economic growth
it is often mistakenly thought to be the reason for why
economics is called the “dismall science”
His key assumption is that technological progress is slow
given the historical record, this was a very reasonable
assumption in 1798 (the irony is that the book was published
just when technological innovation was about to explode)
While the Malthusian model does poorly in explaining the
period following its publication, it remains the workhorse
model for explaining pre-Industrial economic growth
Matti Sarvimäki
Economic History
1: Introduction
18 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Testing the Malthusian model: The Plague
The Malthusian model suggests that living standards are
stagnant in the long-term
in the “short-run” per capita income should fluctuate with
population (population ↓ → wages ↑)
The canonial example
the Plague killed between one third and one half of the
European population in 1348–50
Matti Sarvimäki
Economic History
1: Introduction
19 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Real wages in England, 1200–1869
Clark (2005, JPE)
working class in england
1311
Fig. 4.—Real wages, 1200–1869, Phelps Brown and Hopkins vs. new series. In both
Note: series,
PBH 1860–69
refers tohasthe
earlier
by Phelps
PhelpsBrown
Brown
Hopkins
(1981)
been
set toestimates
100. Sources:
andand
Hopkins
(1981,
28–31),
Matti Sarvimäkitable A2.
Economic History
1: Introduction
20 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Real wages vs. Population in England, 1280–1869
Clark (2005, JPE)
1312
journal of political economy
Fig. 5.—Real wages vs. population on the new series, 1280s–1860s. The line summarizing
the trade-off
population and
the preindustrial
era is fitted
using
“[...] the break
frombetween
the Malthusian
erareal
ofwages
littleforadvance
in efficiency
in England
the data from 1260–69 to 1590–99. Sources: population, same as for fig. 3; real wage,
began circa
1640,
table
A2. long before the famous Industrial Revolution, and before even the
emergence of the modern political regime in England in 1689”
Matti Sarvimäki
Economic History
1: Introduction
21 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Testing the Malthusian model: Poor Relief
Boyer (1989, JPE)
Child allowances were commonly paid to laborers with large
families in the early 19th century England
Malthus: allowances cause the birth rate to increase
Matti Sarvimäki
Economic History
1: Introduction
22 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Testing the Malthusian model: Poor Relief
Boyer (1989, JPE)
Child allowances were commonly paid to laborers with large
families in the early 19th century England
Malthus: allowances cause the birth rate to increase
Boyer (1989) argues that Malthus was right
parish level data for 1826–30 (N=214)
birth rates positively associated with child allowances
(conditional on income, population density, child mortality, housing
availability, presence of allotments of farmland to laborers, presence of
cottage industry, agricultural employment share and distance to London)
Question
what needs to be true for Boyer’s estimates to be causal?
Matti Sarvimäki
Economic History
1: Introduction
22 / 37
Spain’s converted Muslims ("Moriscos") were given three days to leave their homes and board ships
destined for foreign lands in 1609.
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Testing the Malthusian model: The Moriscos
Chaney, Hornbeck (2014)
In 1609, Spain suddenly and unexpectedly expelled roughly
300,000 Moriscos (converted Muslims)
In the Kingdom of Valencia, 1/3 of the population was expelled
Matti Sarvimäki
Economic History
1: Introduction
23 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Testing the Malthusian model: The Moriscos
Chaney, Hornbeck (2014)
In 1609, Spain suddenly and unexpectedly expelled roughly
300,000 Moriscos (converted Muslims)
In the Kingdom of Valencia, 1/3 of the population was expelled
CH examine the impact of the expulsion to later population
growth and agricultural output in Valencia and find that it
decreased total output, increased per capita output
increased the persistence of extractive institutions
Matti Sarvimäki
Economic History
1: Introduction
23 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Background
Chaney, Hornbeck (2014)
711: Islamic forces invaded the Iberian Peninsula
many areas prospered economically under the Muslim rule
Valencia became one of the largest cities in Western Europe
Matti Sarvimäki
Economic History
1: Introduction
24 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Background
Chaney, Hornbeck (2014)
711: Islamic forces invaded the Iberian Peninsula
many areas prospered economically under the Muslim rule
Valencia became one of the largest cities in Western Europe
1238: Valencia taken over by the Christians
much of the land and legal jurisdiction ceded to the nobility
Muslims forced to provide labor services, large share of harvests
Christians granted relatively favorable conditions
Matti Sarvimäki
Economic History
1: Introduction
24 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Background
Chaney, Hornbeck (2014)
711: Islamic forces invaded the Iberian Peninsula
many areas prospered economically under the Muslim rule
Valencia became one of the largest cities in Western Europe
1238: Valencia taken over by the Christians
much of the land and legal jurisdiction ceded to the nobility
Muslims forced to provide labor services, large share of harvests
Christians granted relatively favorable conditions
1525: forced conversion to Christianity
the former-Muslim population became known as Moriscos, who
“lived like Christians and payed like Muslims”
nobility resisted growing pressure to expel the Moriscos
(according to a popular saying: quien tiene moro, tiene oro)
Matti Sarvimäki
Economic History
1: Introduction
24 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Expulsion
Chaney, Hornbeck (2014)
Expulsion ordered in Sept. 1609
approximately 130,000 Moriscos (1/3 of population) exported
Moriscos allowed to take only what they could carry with them
Matti Sarvimäki
Economic History
1: Introduction
25 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Expulsion
Chaney, Hornbeck (2014)
Expulsion ordered in Sept. 1609
approximately 130,000 Moriscos (1/3 of population) exported
Moriscos allowed to take only what they could carry with them
Immediate impact
many nobles bankrupted, creditors faced financial ruin
Valencia’s central bank collapsed in 1613
Matti Sarvimäki
Economic History
1: Introduction
25 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Expulsion
Chaney, Hornbeck (2014)
Expulsion ordered in Sept. 1609
approximately 130,000 Moriscos (1/3 of population) exported
Moriscos allowed to take only what they could carry with them
Immediate impact
many nobles bankrupted, creditors faced financial ruin
Valencia’s central bank collapsed in 1613
The Crown worked out a “rescue package”
creditors took substantial losses, nobility agreed to pay a share
nobility allowed to impose extractive taxes on Christians
Matti Sarvimäki
Economic History
1: Introduction
25 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Christian settlement
Chaney, Hornbeck (2014)
Initially belief that Christians would face the same (low) taxes
as elsewhere → migrants began to arrive immediately
Once they realized that taxes would be high, many left
the Crown issued edicts that limited outmigration
temporary tax concessions to attract migrants ... often in the
form of debt that further limited outmigration
1693: (failed) Peasant Revolt
An aside: after the Plague of 1349, England passed the Statue
of Laborers in 1351, which was quite similar as the laws issued
in Vallencia ... and also resulted a (more successful) Peasant
Revolt in 1381
1808: Napoleon declared the abolition of the Spanish
seignorial regime
Matti Sarvimäki
Economic History
1: Introduction
26 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Data
Chaney, Hornbeck (2014)
Data from tithing auctions
the Archbishopric entitled to about 10% of agricultural output
rather than directly collect agricultural goods, the
Archbishopric auctioned the right to collect its share
winning bid value by district and time are available
Augmented with population data
historical maps: geographic area of each tithing district
population data: town recods for 1569, 1609, 1622, 1646,
1692, 1712, 1730, 1768, and 1786
Matti Sarvimäki
Economic History
1: Introduction
27 / 37
Morisco Population Share in 1609
Chaney, Hornbeck (2014)
!"#$%&'()''*+,-.&'/"01%"2103'*4+5&5'67'89%"029':9-$.+1"9;'*4+%&'";'(<=>'
0% Morisco in 31
districts (white)
0–100% Morisco in 36
districts (light gray)
100% Morisco in 31
districts (medium gray)
In the second category,
Morisco communities
were largely segregated
and their average
population share was
50%.
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Baseline District Characteristics
Chaney, Hornbeck (2014)
Table 1. Baseline District Characteristics, by Morisco Population Share in 1609
Log Difference by 1609 Morisco Share:
Pre-Expulsion
Sample Mean
Basic Difference
With Geographic Controls
District Outcome in 1609:
(1)
(2)
(3)
Population per km2
9.13
- 0.496**
- 0.229
[17.6]
(0.181)
(0.173)
Output per km2
379
- 1.218**
- 0.736**
[1256]
(0.238)
(0.229)
Output per capita
30.3
- 0.695**
- 0.508**
[19.4]
(0.127)
(0.125)
Notes: Column (1) reports average district characteristics in 1609 (in levels), and the standard deviation is
reported in brackets. Output is measured as the auctioned tithe value (in lliures), multiplied by ten. Column (2)
reports the basic
difference
for each
district characteristic
(in logs)
the district's
Morisco population
Estimates
from
regressing
district-level
outcomes
inby1609
on districts’
Morisco share in
1609: the coefficients
are1609.
estimated
by regressing
the indicated
characteristic
on theand
district's
Morisco
population
share in
Column
2 reports
that county
population,
output,
output
per
population
in 1609 (between
0 and
Column districts
(3) reports in
the 1609.
estimated
difference
whenobserved
controlling also
capita
areshare
substantially
lower
in 1).
Morisco
Part
of these
for seven geographic characteristics of districts (distance to the City of Valencia, distance to the coast, terrain
differences
can be attributed to observed geographic differences between Morisco and
ruggedness, latitude and longitude, average rainfall, and agricultural suitability). There are 98 districts with
Christian
districts (distance to the City of Valencia, distance to the coast, average
population data in 1609 and 96 districts with output data in 1609. Robust standard errors are reported in
terrain
slope,
latitude,
longitude,
average
agricultural
suitability).
Lower
parentheses:
** denotes
statistical
significance
at 1%, rainfall,
* denotes statistical
significance
at 5%.
measured “output” in Morisco districts may partly be a statistical artifact, reflecting
the historical management of Church tithes.
Matti Sarvimäki
Economic History
1: Introduction
29 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Preliminary analysis
Chaney, Hornbeck (2014)
Differences-in-differences approach
compare changes in districts hardest hit to changes in disricts
that were less affected
Identication assumption
districts with a greater Morisco population share in 1609 would
have changed similarly as the other districts, if not for the
Moriscos’ expulsion (conditional on geographic characteristics)
Matti Sarvimäki
Economic History
1: Introduction
30 / 37
Changes in log population
Chaney, Hornbeck (2014)
!"#$%&'()''*+,"-.,&/'01.2#&+'"2'3"+,%"4,'5$,46-&+7'89':;<='>6%"+46'?6@$A.,"62'B1.%&'
!"#$%&'(&&)*+&!*,-%"./*#&
!"#$%&0(&&)*+&1-.,-.&23/.4$56&
Estimates for βt from regression lnPdt = βt Moriscod + αt + αd + γt Xd + dt , where
Pdt is the population of district d in year t, Moriscod is the population share of
Moriscos in 1609, year fixed-effects αt capture changes common to all districts,
district fixed effects αd caputure time-invariant unobservable differences across
districts and Xd is a vector of geographic characteristics. The solid circles indicate the
point estimates in each year, relative to the omitted base year of 1609, and the
vertical lines indicate 95% confidence intervals.
&
Changes in log output, output per capita
Chaney, Hornbeck (2014)
!"#$%&0(&&)*+&1-.,-.&23/.4$56&
!"#$%&7(&&)*+&1-.,-.&23/.4$56&,$8&9",/."&
&
&
&
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Preliminary analysis: discussion
Chaney, Hornbeck (2014)
Former-Morisco districts experienced
relative declines in population and output
relative increase in output per capita
(population remained persistently lower, but output recovered)
Matti Sarvimäki
Economic History
1: Introduction
33 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Preliminary analysis: discussion
Chaney, Hornbeck (2014)
Former-Morisco districts experienced
relative declines in population and output
relative increase in output per capita
(population remained persistently lower, but output recovered)
Limitations
Malthusian theory implies an explicit dynamic relationship
(districts’ growth rate depends on their initial outcome value)
Morisco and Christian different already prior to the expulsion
→ they should experience differential growth
difficult to quantify whether convergence in former-Morisco
districts was slower than might be expected
Matti Sarvimäki
Economic History
1: Introduction
33 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Preliminary analysis: discussion
Chaney, Hornbeck (2014)
Former-Morisco districts experienced
relative declines in population and output
relative increase in output per capita
(population remained persistently lower, but output recovered)
Limitations
Malthusian theory implies an explicit dynamic relationship
(districts’ growth rate depends on their initial outcome value)
Morisco and Christian different already prior to the expulsion
→ they should experience differential growth
difficult to quantify whether convergence in former-Morisco
districts was slower than might be expected
PhD students: pay attention to how CH discuss the
magnitudes of the estimates (page 14, last paragraph)
Matti Sarvimäki
Economic History
1: Introduction
33 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Estimating convergence
Chaney, Hornbeck (2014)
Estimation equation
lnYd,t − lnYd,t−τ
= δlnYd,t−τ + αt + αd + γt Xd + βt Moriscod + dt
τ
where Moriscod is the population share of Moriscos in 1609, year fixed-effects αt
capture changes common to all districts, district fixed effects αd caputure
time-invariant unobservable differences across districts and Xd is a vector of
geographic characteristics.
Districts modeled as converging to steady-state
annual growth rate depends on the difference between
steady-state and true outcome in period t − τ
steady-state unobserved, but allowed to vary by district, year
and geographic characteristics
Econometric challenges: see the paper
Matti Sarvimäki
Economic History
1: Introduction
34 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Results
Chaney, Hornbeck (2014)
Population growth in former-Morisco areas
substantially lower in 1609–1622
...and remained so until 1786
note that they should grow faster after loosing population
Output growth in former-Morisco areas
substantially lower in 1609–1622
...and remained so for some time (mostly converging along the
expected growth path through the 18th century)
Output per capita in former-Morisco areas
substantially lower in 1609–1622
...and remained so until 1786
Matti Sarvimäki
Economic History
1: Introduction
35 / 37
Introduction
Course logistics
Malthusian model
Moriscos
Papers for the essays
Conclusions
Chaney, Hornbeck (2014)
Extractive institutions
persisted despite increased labor scarcity
an example how labor may empower workers, but also
encourage elites to strengthen efforts to coerce them
slowed population convergence by limiting labor income
Do the results invalidate the Malthusian model?
no, if disposable income did not rise with output per capita
also: estimates for the entire sample region indicate generally
fast convergence in population and output
Matti Sarvimäki
Economic History
1: Introduction
36 / 37
Papers for the essays
Young (2005): The Gift of the Dying: The Tragedy of AIDS
and the Welfare of Future African Generations, QJE 120:
423-466
This paper simulates the impact of the AIDS epidemic using a
model, where the epidemic affects (a) human capital
accumulation and (b) fertility. It argues that the epidemic, on
net, enhances the future per capita consumption possibilities of
the South African economy.
Voigtländer, Voth (2013): How the West "Invented" Fertility
Restriction. AER, 103(6): 2227-64
This paper argues that the Plague in 1348–1350 improved
female labor market prospects by triggering a shift towards the
pastoral sector. As a consequence, the European Marriage
Pattern (later marriage) emerged and reduced childbirths by
approximately one-third.
Appendix
Appendix
Causal questions
OLS
Dif-in-Dif
Avoiding past mistakes
“Our basic message is simple: We have been here before. [...]
Recognizing these analogies and precedents is an essential step
toward improving our global financial system, both to reduce the
risk of future crisis and to better handle catastrophes when they
happen.” (Reinhart & Rogoff: This Time Is Different: Eight Centuries of Financial
Folly, 2009, Introduction)
Matti Sarvimäki
Economic History
1: Introduction
39 / 37
Appendix
Causal questions
OLS
Dif-in-Dif
Testing formal models
A design of experiments [...] is an essential appendix to any quantitative
theory. [Experiments] may be grouped into two different classes, namely,
(1) experiments that we should like to make to see if certain real
economic phenomena—when artificially isolated from "other
influences"—would verify certain hypotheses, and (2) the stream of
experiments that Nature is steadily turning out from her own enormous
laboratory, and which we merely watch as passive observers. (Haavelmo:
The Probability Approach in Econometrics, 1944, Econometrica, 12:1–115)
Matti Sarvimäki
Economic History
1: Introduction
40 / 37
Appendix
Causal questions
OLS
Dif-in-Dif
Explaining Economic Growth
This Course
“...the factors we have listed (innovation, economies of scale,
education, capital accumulation etc.) are not causes of growth;
they are growth.” (North and Thomas: The Rise of the Western World: A New
Economic History, 1973, p. 2)
“There must be some other reasons underneath those, reasons
which we will refer to as fundamental causes of economic growth.
It is these reasons that are preventing many countries from
investing enough in technology, physical capital and human
capital.” (Acemoglu: Introduction to Modern Economic Growth, 2007, Ch. 4)
Matti Sarvimäki
Economic History
1: Introduction
41 / 37
Appendix
Causal questions
OLS
Dif-in-Dif
The English Clergymen
Malthus was one of the English clergymen, who were beneficiaries of “a
system that provided [them] with an extremely good living and required
little in return [...] Though no one intended it, the effect was to create a
class of well-educated, wealthy people who had immense amounts of time
on their hands. In consequence many of them began, quite
spontaneously, to do remarkable things” (Bryson 2010, 34–36)
Other examples include
Edmund Cartwrigth, inventor of the power loom
George Bayldon, complier of the first dictionary of Icelandic
Laurence Sterne, author of popular novels
Jack Russell, “inventor” of the Russell terrier
William Greenwell, founding father of modern archeaology
Matti Sarvimäki
Economic History
1: Introduction
42 / 37
Steady States in a Model with Multiple Equilibria
Voigtländer and Voth (2013, REStud)
REVIEW OF ECONOMIC STUDIES
birth rate
Steady states with "Horsemen effect"
Death/Birth rate
he standard Malthusian model
birth rate
EH
E
0
EU
death rate
death rate
Wage
w0
w
H
Downloaded from http://restud.o
Figure 1
Malthusian
modelmodel
where
rates effect”
increase
eady statesA
in the
standard Malthusian
and death
with “Horsemen
Wage
with income over some range.
Steady state E0 combines low income per capita with low mortality, while
steady state EH is characterized by higher wages and higher mortality. Point EU
an unstable
thatafter
the the
economy
starts out in E0 . A
. This is isequivalent
to a steady
“ratchetstate.
effect”:Suppose
Wage gains
Black Death
major positive shock to wages (beyond EU ) will trigger a transition to the
6
higher-income steady state. VV argue that higher income may have increased
ture to obtain multiple steady states in Figure 1 is an upward-sloping part
death rates in Europe by increasing urbanization, trade and conflict.
ule. This reflects the historical realities of early modern Europe.7 According
factors reducing population pressure include “vicious customs with respect
Appendix
Causal questions
OLS
Dif-in-Dif
Causal questions
We are typically interested in questions suchs as the impact of
increase in living standards on fertility, mortality
culture, institutions etc. on economic growth
These are causal question, i.e. they concern differences
between counterfactual states of the world
Matti Sarvimäki
Economic History
1: Introduction
44 / 37
Potential Outcomes
Easiest to think in the context of a binary "treatment"
(
y1i if Di = 1
potential outcome =
y0i if Di = 0
y1i =outcome for unit i if he/she/it receives the “treatment” Di ; y0i =outcome for the
same unit if he/she/it does not receive the treatment
Potential Outcomes
Easiest to think in the context of a binary "treatment"
(
y1i if Di = 1
potential outcome =
y0i if Di = 0
y1i =outcome for unit i if he/she/it receives the “treatment” Di ; y0i =outcome for the
same unit if he/she/it does not receive the treatment
The causal effect (for unit i)
[y1i − y0i ]
the difference in the potential outcomes with and without the treatment for unit i
Potential Outcomes
Easiest to think in the context of a binary "treatment"
(
y1i if Di = 1
potential outcome =
y0i if Di = 0
y1i =outcome for unit i if he/she/it receives the “treatment” Di ; y0i =outcome for the
same unit if he/she/it does not receive the treatment
The causal effect (for unit i)
[y1i − y0i ]
the difference in the potential outcomes with and without the treatment for unit i
But we only observe
yi = yoi + [y1i − y0i ] Di
Identification
The aim is to construct a plausible counterfactual
“reduced form”: a control group that tells us of what would
have happened to the treatment group without the treatment
“structural models”: estimate parameters of a model that can
then be used to predict counterfactual outcomes
Identification
The aim is to construct a plausible counterfactual
“reduced form”: a control group that tells us of what would
have happened to the treatment group without the treatment
“structural models”: estimate parameters of a model that can
then be used to predict counterfactual outcomes
The treatment effect is identified if
a valid control group exists
a model is a good approximation of reality and a valid control
group for identifying its parameters exist
Identification
The aim is to construct a plausible counterfactual
“reduced form”: a control group that tells us of what would
have happened to the treatment group without the treatment
“structural models”: estimate parameters of a model that can
then be used to predict counterfactual outcomes
The treatment effect is identified if
a valid control group exists
a model is a good approximation of reality and a valid control
group for identifying its parameters exist
Selection bias
[what would have happened to the treatment group without
the treatment] − [what happened to the control group]
Appendix
Causal questions
OLS
Dif-in-Dif
Point Estimates and Standard Errors
Most of the time, we are interested in two numbers
Point estimate
Standard error
Matti Sarvimäki
Economic History
1: Introduction
47 / 37
Appendix
Causal questions
OLS
Dif-in-Dif
Point Estimates and Standard Errors
Most of the time, we are interested in two numbers
Point estimate
Standard error
An estimate is statistically significant when it is at least two
standard errors away from zero
Matti Sarvimäki
Economic History
1: Introduction
47 / 37
Appendix
Causal questions
OLS
Dif-in-Dif
Point Estimates and Standard Errors
Most of the time, we are interested in two numbers
Point estimate
Standard error
An estimate is statistically significant when it is at least two
standard errors away from zero
An estimate may be statistically insignificant because
precisely estimated point estimate is close to zero
standard errors are large (i.e. sample size is too small)
→ this does not prove that the true effect is zero/small.
Matti Sarvimäki
Economic History
1: Introduction
47 / 37
Appendix
Causal questions
OLS
Dif-in-Dif
Point Estimates and Standard Errors
Most of the time, we are interested in two numbers
Point estimate
Standard error
An estimate is statistically significant when it is at least two
standard errors away from zero
An estimate may be statistically insignificant because
precisely estimated point estimate is close to zero
standard errors are large (i.e. sample size is too small)
→ this does not prove that the true effect is zero/small.
Beware of underpowered studies! Read Gelman, Weakliem
Of Beauty, Sex and Power, American Scientist 97
(2009):
Matti Sarvimäki
Economic History
1: Introduction
47 / 37
Ordinary Least Squares (OLS)
Estimation equation
yi = αDi + Xi β + i
yi = outcome, Di = treatment (0/1 or “exposure”), Xi = observable factors, i =
unobservable factors.
An estimate of α is a
(weighted) average difference in y
... between units that have different treatments
... but identical observed characteristics
Ordinary Least Squares (OLS)
Estimation equation
yi = αDi + Xi β + i
yi = outcome, Di = treatment (0/1 or “exposure”), Xi = observable factors, i =
unobservable factors.
An estimate of α is a
(weighted) average difference in y
... between units that have different treatments
... but identical observed characteristics
This is a causal effect if
Di and i are conditionally independent: for units with
identical Xi , treatment is as good as randomly assigned
Appendix
Causal questions
OLS
Dif-in-Dif
Differences-in-Differences (Dif-in-Dif or DD)
Dif-in-dif is based on the assumption that the counterfactual
trend in the treatment and control groups are the same
i.e. there may be permanent unobserved differences between
the treated and untreated as long as counterfactual changes
are paraller
Matti Sarvimäki
Economic History
1: Introduction
49 / 37
Appendix
Causal questions
OLS
Dif-in-Dif
Differences-in-Differences (Dif-in-Dif or DD)
Dif-in-dif is based on the assumption that the counterfactual
trend in the treatment and control groups are the same
i.e. there may be permanent unobserved differences between
the treated and untreated as long as counterfactual changes
are paraller
The most basic dif-in-dif setup looks like this
Post-Period
b
d
Dif-in-Dif
where a,b,c and d are means of the outcome
Treatment Group
Control Group
Matti Sarvimäki
Pre-Period
a
c
Economic History
Dif
b-a
d-c
(b-a)-(d-c)
1: Introduction
49 / 37
How does Dif-in-Dif work?
Suppose that the true data generating process is
yit = α (Dit × postt ) + βpostt + θDi + it
where yit is the outcome, Dit treatment status (0/1), postt is a dummy for the
post-peridod (they captures general changes between pre- and post-period that affect
both the treatment and the control group), Di captures time-invariant differences
between the treament and control group, and it are unit specific time-variant
unobserved factors. That is, the treatment group differs from the control group
(regarless of the treatment) by the amount θ.
How does Dif-in-Dif work?
Suppose that the true data generating process is
yit = α (Dit × postt ) + βpostt + θDi + it
where yit is the outcome, Dit treatment status (0/1), postt is a dummy for the
post-peridod (they captures general changes between pre- and post-period that affect
both the treatment and the control group), Di captures time-invariant differences
between the treament and control group, and it are unit specific time-variant
unobserved factors. That is, the treatment group differs from the control group
(regarless of the treatment) by the amount θ.
Dif-in-Dif
Period 0: no-one treated. Period 1: treatment group (T=1) treated
h i
h i
E y1T − E y0T
h
E
Even if θ 6= 0, we have E
=
α+β
=
β
i
y1T − y0T − y1C − y0C = α
y0C
i
(α + β + θ) − (θ)
y1C
−E
h
=
Generalized Dif-in-Dif
Estimation equation
yit = αt (Di × δt ) + δt + ηi + it
yit is the outcome, Di treatment status (0/1 or “exposure”), δt is time-variant
unobserved factor that affects everyone, and ηi and it are unit specific time-invariant
and time-variant unobserved factors. Note that α now has a time subscript t.
Now there may be several pre- and post periods (can estimate
the dynamics of the effects)
αt measure the differences in changes (in comparison to some
baseline year) between the treatment and control groups
Note that the “main effect” of Di is captured by vector ηi , but
the interactions between treatment and year remains identified