Time Series

GENUS, LXVII (No. 2), 119-139

TOMMY BENGTSSON* - GÖRAN BROSTRÖM**

Famines and mortality crises in 18th to 19th century
southern Sweden

1. INTRODUCTION

Analyses of causes of variations in mortality have basically followed two
paths. The first is focusing on years of mortality crises. One question is
whether or not these crises are caused by starvation, inflicted by bad harvests
orresult from market failure. Sen’s analyses of the famines in Bengal,
Ethiopia, Sahel, and Bangladesh are examples of this approach (Sen, 1981,
2001). Analyses of European famines from the sixteenth century onwards (see
Walter and Schofield, 1989) and of the situation in Warsaw and Amsterdam
during the Second World War are other examples (see Livi Bacci, 1991). Typical of these studies is thefocus on extreme situations. Almost as a rule, only
years of excess mortality are selected. Ó Gráda’s recent overview of famines
all over the world proves the case: out of the twenty-four famines selected,
from France in 1693-1694 to Malawi in 2002 and Niger in 2005, all but the
latter two are associated with excess mortality (Ó Gráda, 2009, Table 1.1).
However, studies focusing on mortalitycrises have short-comings in their
analyses of causes since is likely to create a selection bias, as described by
Heckman (2008). Their merits are instead explorative and in the analyses of
consequences.
The second path takes a time series approach, analysing causes of variation in mortality over a certain period of time, often a hundred years or more
(for an overview, see Bengtsson andReher,1998). The focus is on testing
hypotheses of causes of mortality variation, such as effects of fluctuations in
real wages, food prices, and temperature. Such factors often have a strong
time trend as well as a fair degree of fluctuation around this trend. Due to the
length of the time periods analysed, the trend in real wages and food prices
itself contains little unique information. Thereason is that other factors, such
as urbanisation and the expansion of schooling and health care, show similar
trends, which makes it difficult to establish causality. The focus has therefore
been on the effects of the fluctuations of food prices, temperature and other
external factors on demographic outcomes. This has proven very useful since
* Centre for Economic Demography (CED) andDepartment of Economic History, Lund University, Sweden.
** Centre for Population Studies (ALC) and Department of Statistics, Umeå University, Sweden.
Corresponding author: Tommy Bengtsson; e-mail: tommy.bengtsson@ekh.lu.se.

119

TOMMY BENGTSSON - GÖRAN BROSTRÖM

it provides a basis to establish causality, as the series after trend removal contains a large amount of variation unique to each ofthe series.
Numerous studies using the time series approach, based on annual aggregated data from preindustrial populations, show that fluctuations in food
prices affect demographic outcomes, particularly fertility, but also mortality
and migration (Bengtsson, 1993a; Galloway, 1988; Lee, 1981). The results
refer not only to preindustrial Europe but also other parts of the world (Lee,
1990;Bengtsson and Reher, 1998).
Most of these studies are based on the total number of events for the
entire population of a certain area, often a country. From an analysis of agespecific mortality for Sweden, we know however, that it was in particular
adults and children aged five years and above that were particularly vulnerable to short-term economic stress (Bengtsson and Ohlsson, 1985, p. 317). Themortality among infants seems to follow its own rhythm, likely a result of
breastfeeding practices (Bengtsson and Ohlsson, 1985, p. 317; Utterström,
1957). Not only fluctuations in food prices influence demographic events;
cold winters and warm summers also affect demographic outcomes (Lee,
1981; Richards, 1984; Tromp, 1963).
These studies are often estimating distributed lag models with...